Read more

On 2 August, Blue Owl hired Takeo Ikemori into its Tokyo office as head of Japan institutional capital. He spent sixteen years at Barclays and ran Natixis’ financial institutions group in Tokyo before that. He shares the Japan representative role with Yoichi Nakamura, who runs private wealth in the country.

I’ve read maybe forty versions of this story over the years and they all say the same thing. Big American manager opens a seat in Tokyo. Asia is the growth market. Everyone’s doing it.

My problem with that reading is that it’s incurious. It never asks which seat, or what the seat is replacing.

The consensus: Japan at 5-15% against 40% elsewhere

Here is the strongest version of the view I’m departing from, and most of it is simply correct.

Japanese institutions are the most under allocated large pool of capital in the developed world. CJ Morrell of Fiera Capital puts Japanese institutional allocations to private markets at “five per cent to 15 per cent” against “around 40 per cent in Canada, the US and Western Europe,” and describes Japan Inc as “still very much stuck in that 60 per cent debt, 30 per cent equity, 10 per cent hedge fund/alternatives.” He calls it “the very early stages of what I think will be a multi-decade move to embrace private markets.”

The mandate side is moving too. Japan is steering its $1.8 trillion public pension fund toward alternatives, which sat at 1.7% of holdings as of March against a 5% ceiling.

So: enormous pool, structurally under weight, mandate pushing the same direction, a decade of runway. Every large alternatives manager should be staffing Tokyo. Blue Owl’s Co-CEOs said as much in the June quarter release, crediting “the diversification of our business across platforms and geographies.”

I sign all of that. I read it as the right description of Japan and an incomplete description of Blue Owl.

Credit fundraising fell 69% while AUM rose 12%

Blue Owl’s June quarter looks fine from the top. AUM reached $319.0 billion, up 12% year over year. Fee-Related Earnings were $392.2 million and Distributable Earnings $351.2 million, both up 9%. The firm declared its $0.23 quarterly dividend.

Now open the fundraising page of the same deck, which is where I think the story sits.

Total equity fundraise was $7.6 billion in the quarter, against $12.1 billion in the June quarter of 2025. That’s a decline of 37% in the rate at which new equity arrives. AUM grew 12% while the intake that produces future AUM fell by more than a third.

Split it by platform and the decline concentrates in one place:

Segment figures are rounded independently to $0.1 billion, so the columns do not foot exactly to the stated totals.

Credit equity fundraising fell 69% year over year, from $5.8 billion to $1.8 billion. Credit is the franchise. It houses the direct lending business, the non traded BDCs, and the brand.

One thing I have to flag before leaning on that number, because it cuts against me: the 2Q’25 base was itself a record quarter for the credit platform. Measuring a soft quarter against a record makes any decline look worse than it would against a trailing average. The fall is real and large on any base I can construct, and it is not as clean as 69% implies.

Split the same quarter by channel and it resolves further. Private wealth equity fundraise was $1.7 billion. Institutional equity fundraise was $5.9 billion. Those add to the $7.6 billion total. On a trailing twelve month basis it runs $13.8 billion private wealth against $26.1 billion institutional.

Institutional money is already carrying roughly three quarters of the intake. I’d note that’s at a firm whose growth story for four years was the wealth channel.

The portfolio is running in place

One more number from the same deck, and I think it’s the most under quoted figure in the release.

Direct lending originations were $3.6 billion in the quarter, and net deployment was $0.6 billion. Over twelve months, originations were $33.2 billion against net deployment of $6.3 billion.

Roughly four fifths of what Blue Owl originates is replacing something that repaid. A lender doing $33 billion of gross origination to grow the book by $6 billion is running hard to stay near the same place. That is not a solvency observation and I don’t want it read as one, because repayments are the healthy way for a credit book to generate liquidity. It’s a growth observation, and it explains where the cash to fill the redemption queue comes from.

The queue, and the arithmetic of 27 cents

Blue Owl Credit Income Corp, the non traded BDC known as OCIC, received $3.6 billion of repurchase requests in the June quarter, down from $4.2 billion in March. Those are 18.8% of shares outstanding, down from 21.9%, against a repurchase cap of 5%.

Work the ratio, because the ratio is the thing.

An investor who asked for a dollar back in the June quarter received about 27 cents. In March they received about 23 cents. OCIC’s own filing publishes exactly these fill rates, “representing 22.8% of each shareholder’s tender request” in the first quarter and “approximately 27%” in the second, so this is the issuer’s arithmetic and not merely mine. The improvement is real and I won’t wave it away. It’s also a move from very gated to gated.

Those two percentages let me back out the fund’s size, which Blue Owl does not disclose directly:

Two quarters, two independently reported percentages, implying the same roughly $19.2 billion of OCIC net asset value to within about 0.15%. A 5% gate on that base is roughly $958 million a quarter of capacity against $3.6 billion of demand.

The number that stopped me is on the other side of the ledger. OCIC has taken in about $1.2 billion of total capital inflows year to date, including roughly $790 million of subscriptions and an estimated $380 million from dividend reinvestment, which the filing itself flags as an estimate. So the fund gathered $1.2 billion across six months while being asked for $3.6 billion back in three. That’s the ratio I keep coming back to.

Blue Owl’s answer to this is worth putting in its own words, because it’s a good one. Quarterly repayments run “typically 6-8% of total assets” against tender offers of 5% of net assets, OCIC has taken $2.7 billion of ordinary course repayments year to date at “approximately 1.4x coverage of share repurchases,” and the fund “does not need to sell a single private loan to satisfy the tender offer.” I accept all of that. It’s an argument about whether the queue is fundable, and my argument is about what the queue does to sales.

This is not a Blue Owl specific failure. Apollo, Ares, Morgan Stanley, HPS, Cliffwater, Monroe and Blackstone all saw requests exceed 5% and held repurchases at the cap. I flagged the structural headwinds in multi strategy funds’ private credit expansion last November, when the strain sat on the manager side rather than the wealth channel. Robert A. Stanger data reported by the Wall Street Journal size it. Investors asked for $15.6 billion back and received $5.9 billion in the June quarter, against $13.9 billion and $7.4 billion in March, while new money into private credit funds fell to around $500 million in May, the smallest inflow in at least eighteen months.

The industry’s fill rate fell from 53% to 38% while Blue Owl’s rose from 23% to 27%. Blue Owl is improving off a base materially worse than its peer group. Oaktree shows the gate can open again: its requests dropped to 4.5% of shares from 8.5%, under the cap, so its queue cleared in full.

On cause, I’d be careful. There is some credit evidence: 6.4% of private credit loans carried “bad PIK” at Q4 2025, interest deferred mid loan under liquidity strain, which Lincoln International treats as a shadow default rate implying distress near 6% against a headline near 2%. But Fitch reads the redemptions as sentiment driven, and the sharper diagnosis in that same analysis is product design: “semi-liquid” was sold as “something meaningfully more accessible than the quarterly cap mechanics permit.” I find that more persuasive, and it matters: a sentiment queue can reverse and a credit queue can’t.

What the gate actually costs, which is less than it looks

Here’s where I have to argue against my own framing, because the obvious next move is to say that 85% of Blue Owl’s fees sit in vehicles with redemption queues, and that would be misleading.

The firm reports Permanent Capital of $225.0 billion and states it generated 85% of FRE management fees over twelve months. Its defined term concedes that permanent capital includes products that “may have periodic tender offers or redemptions.” I thought that was a revelation until I checked the peers. Blackstone, Apollo, Ares and KKR all write functionally the same caveat into their own permanent capital definitions. Blackstone’s 10-K carves out vehicles “where required redemptions are limited in quantum, such as interval funds.” So I read it as industry standard language, and I think anyone selling it as a gotcha is overreaching. I nearly did.

So I sized the actual exposure, which is the part I haven’t seen anyone do. OCIC’s roughly $19.2 billion is about 8.5% of the $225.0 billion permanent capital base. A full quarter at the 5% cap removes about $958 million of NAV. OCIC’s base management fee is 1.25% of net assets annually, charged on equity and not on leverage, so that removes about $12 million of annualised fee income, against $392.2 million of Fee-Related Earnings in the quarter.

So the gate is not an earnings event. Two or three basis points of FRE is noise, and redeemed NAV is partly replaced by fresh origination anyway. What the queue actually does is kill the sales channel. My reasoning is that no adviser puts a client into a vehicle that returned 27 cents on the dollar last quarter, and I’d expect that to show up exactly where it did: $1.8 billion of Credit fundraising where there used to be $5.8 billion. I want to be clear that this step is inference. I have the gate and I have the fundraising collapse, and I do not have an adviser survey sitting between them. The damage is to the growth rate, and the growth rate is what a manager’s multiple is made of.

I’ve written the tradeable half of this argument separately: the Blue Owl gate note on Patreon prices the gate against the fee base and works the October filing sequence the way a desk would.

That reframing is what makes the Tokyo hire legible. Blue Owl doesn’t need Japan to defend this year’s fees. It needs Japan to replace a distribution engine.

The timing lever, which is a rate differential

Japanese institutions finally have yield at home. The 10 year JGB is trading near 2.78%, having touched 2.901%, levels last seen in 1996, up more than 70 basis points this year. The Bank of Japan’s policy rate is 1.00% after a June hike. For the first time in a generation a Japanese institution can buy duration domestically with no credit risk and no hedging cost. On that evidence, funding an American direct lender looks like the worse trade.

The binding constraint on Japanese outbound credit was never the JGB yield. It was the cost of hedging the dollar back to yen, which runs at roughly the short rate differential between the two currencies.

The hedge cost proxy has more than halved, because the Fed cut toward 3.50-3.75% and the BOJ hiked at the same time, from both ends at once.

Two limits on that calculation, and the second is the serious one. The short rate differential is a proxy: real hedged returns also carry the cross currency basis, which is persistently negative for yen and which I can’t observe precisely from public data, so the true pickup is somewhat worse than the proxy implies. And rising JGB yields raise the domestic bar in the same window that hedge costs fall. Both effects are live simultaneously, and the article’s claim is only that the second moved further than the first.

Why Yamamoto’s exit is the weaker of my two discriminators

The rival explanation fits most of these facts. It’s an industry land grab, and the case is stronger than I’d like: Apollo is expanding its Asia private wealth staffing after raising close to $5 billion from wealthy Asian investors, and Ares closed a $2.4 billion Japanese data centre fund in 2025. Tokyo in 2026 is crowded with people making this trip. Blue Owl also launched an insurance solutions platform in 2024 and hired a leader for it in May 2026, so a firm wide institutional pivot predates this hire by two years.

Two things still discriminate, and neither is decisive alone.

Blue Owl is backfilling. Takeshi Yamamoto was a managing director at Blue Owl Japan. In April he left to become head of capital formation for Japan at Benefit Street Partners, Franklin Templeton’s private credit arm, where he now argues publicly that “private credit is at a clear inflection point” and that Japanese institutional investors “have been under allocated compared to their global peers.” I’d weigh this less heavily than I first wanted to. Senior Tokyo coverage bankers are being poached across the sector right now, and I think a four month search for an outside hire is ordinary.

The seat is institutional and separate. Nakamura runs private wealth, and has since he joined from Algebris. Ikemori’s book is banks and insurers: he led coverage of Japan’s largest financial institutions at Barclays and ran financial institutions at Natixis Tokyo. Splitting Japan into a wealth seat and an institutional seat, and filling the institutional one with a balance sheet banker, is a statement about which pocket the firm now underwrites.

Four things I can’t rule out

I cannot prove the Credit collapse is a wealth channel story. Blue Owl publishes fundraising by platform and fundraising by channel, and it does not publish the intersection. So I can show that Credit intake fell 69% and that wealth intake is small, and I cannot show from the disclosures that the fall happened in the wealth channel. Institutional pullback from direct lending would fit the same two tables. This is the weakest joint in the argument and I’d rather name it than let a reader find it.

The money may not be fungible into the hole. This is the objection I’d lead with if I were arguing against myself. Nothing guarantees Japanese institutional dollars land in the direct lending sleeve that’s bleeding. Ikemori’s stated mandate spans private credit, real estate and GP stakes, and Blue Owl’s institutional inflows this quarter went to net lease, real estate credit, investment grade credit, direct lending and GP minority stakes, which is a mix. Japan could diversify the institutional base without ever touching OCIC’s problem.

Global LPs are rotating away from exactly this product. A 2026 Rede Partners survey found roughly 70% of institutional LPs expect to diversify beyond direct lending, with only a small minority planning to increase mid market direct lending allocations, citing spread compression and underwriting standards. That is a structural headwind to the institutional bid, and it is independent of anything Blue Owl did.

Japanese money is currently going the other way. Japanese investors were net sellers of foreign bonds on a large scale in early 2026 as domestic yields rose, and GPIF has signalled its alternatives build will tilt toward domestic projects. I take that seriously: the repatriation trend is real, and it cuts directly against a story about Japanese capital funding American credit.

The causal arrow may point the other way. Japan may be the next market on a plan drawn in 2023 when the Tokyo office opened, with the funding stress a coincidence of timing. I can’t see the internal calendar. And the clocks don’t match: a relationship led institutional build in Japan takes years, while the fundraising hole is quarterly.

What would change this view: the late-October print

Blue Owl reports September quarter results in late October 2026, with OCIC’s tender result in the same window.

I’m wrong if Credit equity fundraising recovers toward its $5.8 billion year ago quarter without institutional or Japanese money doing the work, and OCIC’s requests fall under the 5% cap so the gate stops binding. That combination says the wealth channel healed on its own and Tokyo was ordinary expansion.

I’d count it confirmed if Credit intake stays near $1.8 billion while institutional share rises again and Blue Owl starts naming Japanese institutions in its capital formation commentary.

The cleanest single tell is OCIC’s subscription line. Watch whether inflows climb back toward the $3.6 billion quarterly exit demand. That gap, not the fee line, is the thing that has to close.

What I’d actually do with this

If you own OWL, stop underwriting the redemption queue as an earnings risk. It isn’t one, and the arithmetic above says so. Underwrite the distribution risk instead: the wealth channel that produced $5.8 billion of Credit equity a year ago produced $1.8 billion last quarter, and no plausible Japanese institutional ramp replaces $4 billion a quarter within a year. The stock fell 68.2% from its January 2025 peak of $25.02 to $7.95 on 2 April 2026, so the market has priced something. I can’t tell yet whether it priced a fee cut that isn’t coming or a growth rate that is actually broken. That’s the question I’d want answered before I sized anything.

If you’re a credit allocator, note that the whole industry’s wealth channel is gated at 5% at once. That makes the marginal buyer of US direct lending paper more likely to be an insurance or balance sheet account than an American wealth platform. Balance sheet money prices differently and holds through drawdowns that would trigger a retail queue.

If you’re competing for the same Japanese capital, the hedge cost window is a rate differential and rate differentials close.

I covered Blue Owl’s OBDC merger termination and its NAV discount in February, when the pressure sat in one vehicle and read as a governance problem. Six months on I read it as a distribution problem, which is a different and larger thing.

So here’s the question I’d put to anyone who underwrites managers for a living. OCIC gathered $1.2 billion in six months while being asked for $3.6 billion back in three. At what ratio does a perpetual vehicle stop being a growth engine and start being a runoff book? I don’t think Blue Owl is near that line. I do think it’s the only line that matters now, and I haven’t seen anyone put a number on it.

📊 The Decision-Grade Version

This piece is complete on its own. The thesis, the evidence, the confounds and the dates that would kill the view are all above. Nothing was held back to sell you a next step.

The Patreon note is a separate piece of work. It takes one decision from inside this story: pricing the redemption gate against the fee base, timed to the October tender print that lands four weeks before Blue Owl reports. It writes that the way a desk would act on it, with the arithmetic shown, the crowding constraint sized, and the outcomes that confirm or kill the read. Written for people who put capital behind a view.

→ Read the Blue Owl gate trade note

→ Or join the Patreon community for every note

Elsewhere: YouTube for the video breakdowns, and LinkedIn for the shorter reads.

Read more

When Citadel took Situational Awareness’s entire public book in a single block before the open on 30 July, four of the AI-infrastructure names at the centre of it rallied 21–28% in a session. The consensus read that as confidence: a credible buyer had stepped in, so the selloff must have been overdone. I don’t think that reading survives the cross-section. A stock the fund never owned rallied just as hard as the ones it did. That’s the whole story.

Read more

Situational Awareness returned 439% in the first half of 2026 and still sold its public book at the bottom, in a week priced for a rate hike that never came. The decision block, the numbers, and what would kill the view:

On Wednesday, July 29, the Federal Open Market Committee voted 9-3 to hold the federal funds rate at 3.50-3.75%. Before Thursday’s open, Situational Awareness LP, the AI-concentrated hedge fund run by former OpenAI researcher Leopold Aschenbrenner, sold its public equities book in a single block, longs and shorts together, to Ken Griffin’s Citadel. How much of it went is itself disputed: the whole book, per SpotGamma’s account of CNBC’s reporting and The Next Web’s of the FT’s; the majority of it, per TechCrunch. On Thursday, the names in that book ripped: SanDisk rose 23.9%, Micron 17.0%, Bloom Energy 26.4%, with the Nasdaq 100 up 3.1%.

The consensus reading of this episode is already settled: a brilliant, over-levered kid met a margin call, and leverage kills. That reading is true and worthless. Every professional already knows leverage kills. What the consensus misses is what the public record shows on two specific points. The fund’s hedges were structurally incapable of protecting it, and the numbers proving that sat in its own filings since May. And the panic that filled its final week was measurably part hedging flow rather than forecast, a distinction visible in real time on two free screens. Neither required inside information. That is the part worth a professional’s time.

Read more

Man Group’s own framing of its first half of 2026 is a diversification story, and at the firm level the framing holds up on the numbers. Robyn Grew told shareholders that “the exceptional net inflows and record AUM we are reporting today are the direct result of deliberate, multi-year investments in the diversification of our business.” AUM hit an all-time high of $253.6 billion, up from $227.6 billion at the end of 2025 — driven by $19.8 billion of investment performance and $7.1 billion of net inflows, which together with a small negative FX/other adjustment reconcile the full move (227.6 + 7.1 + 19.8 − 0.9 = 253.6). Net inflows ran 3.4 percentage points ahead of the industry — Man Group’s own asset-weighted calculation against HFR/Morningstar/Citi data, which by its own footnote excludes several AUM categories (infrastructure managed accounts, real estate, US private credit, CLOs), not an independently audited industry statistic. Flows were spread across long-only credit ($4.8bn), long-only equity ($1.5bn), private markets ($0.7bn) and liquid alternatives ($0.1bn). Bloomberg’s coverage led with the record-AUM framing, and the market agreed: shares rose 5.7% to 317 pence on the results.

A sophisticated reader already knows large multi-strategy managers diversify across products; that fact alone isn’t the article. The specific, testable claim underneath it is narrower: one of those product lines — systematic long-only, the quantitative equity business run through Man Numeric — has a net-flow number that has been dominated by single institutional decisions for three consecutive reporting periods, in both directions, and that changes what the category’s headline growth or contraction figure actually tells an outside reader.

Read more

EquiLibre Technologies’ $500M venture valuation prices its AI trading operation using software-scaling logic — but Grinold’s law, the 2007 quant unwind, and Renaissance’s own three-decade capacity discipline all show a trading edge scales in the opposite direction from a software product.

Read more

Global hedge fund capital hit a record $5.6 trillion in Q2 2026 on the largest quarterly increase HFR has ever recorded — built in part on gross leverage near eight times NAV, concentrated in a handful of AI-chip names. And when that trade broke in mid-July, it was not a clean leverage unwind: at least four distinct triggers landed in the same window, and financing costs were only one of them.

Read more

Susquehanna Securities says unidentified traders turned $12 million into more than $100 million by front-running a Chinese regulatory crackdown, and it filed a 100-defendant “John Doe” suit in Manhattan federal court to find out who they are. Citadel Securities has since asked to join that suit, saying it lost $28 million more on the same trades — and said so explicitly because Susquehanna’s recovery could shrink the pool left for its own. The consensus read — echoed across Bloomberg, Reuters, and the trade press — is that this is a blockbuster insider-trading bust that the SEC and DOJ will now run down. It probably is that. What it is much less likely to be is a way for either firm to get its money back. The statute they sued under caps and offsets private recovery, the likely defendants sit in a jurisdiction that does not reliably enforce U.S. judgments, and the specific channel the money moved through — brokerage platforms owned by the same companies whose confidential regulatory dealings leaked — is a structural blind spot that pre-dates this case and will outlast it. Six weeks in, as of this writing, that prediction is already visible in the public record: the case has gone quiet exactly where it would need to move for either firm to see a dollar back. For any desk pricing options in single-name China ADRs, the lesson isn’t “insider trading got caught.” It’s that getting caught and getting paid back are two different processes, running on two different clocks, and the market maker eats the loss on the first one regardless of how the second one ends.

What the trading actually looked like

On May 22, 2026, at 4:33 a.m. Eastern time, Reuters reported that China was cracking down on “illegal” cross-border securities activity, naming Futu Holdings, UP Fintech (Tiger Brokers), and Longbridge as targets. Both Futu and UP confirmed the same morning that they’d received China Securities Regulatory Commission enforcement notices — Futu’s filed with the SEC at 9:35 a.m. Eastern, UP’s at 9:16 a.m. The CSRC ultimately proposed to confiscate roughly RMB1.85 billion (about $271 million) in illegal gains and fines from Futu’s mainland and Hong Kong entities, and roughly RMB411 million from UP’s, covering both administrative penalties and confiscated income (Susquehanna complaint; Bloomberg). FUTU opened that day at $81.08, down 34.5% from the prior close of $123.86; TIGR opened at roughly $4.01, down 31.3% from $5.84 (Susquehanna complaint). By the close, the stocks had partly recovered from their lows, finishing down 27.53% at $89.76 (FUTU) and 25.34% at $4.36 (TIGR) (BigGo Finance, corroborating South China Morning Post).

Everything that follows about the trading itself comes from one source: Susquehanna’s own complaint, an adversarial pleading that hasn’t been tested in court and was written by the party claiming the loss. It should be read as Susquehanna’s account, not adjudicated fact — though the underlying numbers (trade counts, strikes, expirations, broker names) are the kind of granular, falsifiable detail a firm would risk sanctions for fabricating, and no defendant has yet appeared to contest them. Susquehanna’s complaint — filed June 29 in the Southern District of New York as Susquehanna Securities, LLC v. John Does 1-100, Case No. 1:26-cv-05474 — lays out that account in unusual granular detail for a civil pleading, because the firm had its own execution data as a market maker on the other side of the trades (complaint). Between May 7 and May 21, Susquehanna alleges, buyers purchased more than 200,000 short-dated put option contracts on FUTU and TIGR — over 50,000 on FUTU, over 150,000 on TIGR — paying roughly $12 million in premium for options that returned more than $100 million, a return in excess of 900%. For scale: Susquehanna’s complaint compares the scheme to Raj Rajaratnam’s Galleon case, citing the $53 million Rajaratnam was criminally ordered to forfeit. That figure is correct for the personal forfeiture, but it understates the broader picture: the government’s own estimate of total profits and losses avoided across the full Galleon scheme, including Rajaratnam’s co-defendants, runs closer to $72 million (SEC Litigation Release No. 21397). Even against that larger number, this case is alleged to be bigger — about 1.4 times Galleon’s total on Susquehanna’s own $100 million-plus estimate, closer to 1.9 times on Citadel’s $137 million estimate — compressed into two weeks instead of years.

The complaint’s evidentiary strength isn’t the size of the return. Announced-catalyst option trades routinely return multiples; that alone proves nothing. What it argues is coordination and specificity. Three brokers — Interactive Brokers, UP’s own TradeUP platform, and Futu’s own trading platform — accounted for about 76.5% of the market-wide purchases matching the suspicious profile, even though the qualifying trades were legal to place through any U.S. broker. At Interactive Brokers specifically, a firm with more than 5 million customer accounts, only about nine accounts were responsible for the relevant purchases. The trades clustered in specific contracts rather than spreading across the options chain: on May 7, the flagged trades made up about 87% of all FUTU near-dated put purchases market-wide; on May 13, about 98% of TIGR’s. And the expirations selected show timing knowledge, not just directional conviction — about 31% of the flagged contracts expired the day after the Crackdown News broke, and another 35% expired the following week, meaning two-thirds of the position was structured to pay off inside roughly seven trading days of a specific, then-unannounced date. Neither company had released negative news in the prior two weeks; Futu’s own quarterly earnings didn’t come until May 28, UP’s until June 2 (complaint).

The mechanic behind that loss is ordinary options math, not anything exotic: Susquehanna and Citadel, as market makers, sold puts and collected the premium upfront, pricing the flow as if it were routine; when FUTU and TIGR gapped down more than 30% on news they didn’t see coming, they owed the buyers the difference between the strike price and the crashed price on every contract — a payout due the moment the options settled in the money, regardless of who eventually turns out to have been on the other side of the trade. Susquehanna’s own exposure: it says it was the seller on contracts where buyers paid $6.7 million in premium and made approximately $71.4 million in profit — hence the $71.4 million floor on the damages it’s seeking (complaint; Bloomberg). The firm won a court order on June 30 authorizing subpoenas to, and freezes on, accounts at Interactive Brokers, TradeUP, and Futu’s platform, to unmask the account holders (Bloomberg). The SEC confirmed by July 2 that it’s reviewing the trades, and the Justice Department’s criminal division has opened its own early-stage inquiry (Philadelphia Inquirer; Yahoo Finance).

Susquehanna was not the only counterparty. On July 2, Citadel Securities filed a motion to join the suit, saying it was “the victim of a brazen insider trader scheme” on the same FUTU and TIGR options and had lost about $28 million as counterparty to trades it, too, did not know were informed (Claims Journal/Bloomberg). Citadel’s own filing puts the total scheme profit at roughly $137 million (Finance Magnates, independently corroborating Claims Journal/Bloomberg) — higher than Susquehanna’s “$100 million-plus” estimate, a discrepancy the two firms’ pleadings don’t reconcile and that is itself a data point: even the two market makers who were on the losing side of most of the flow don’t agree on the size of what hit them, which says something about how fragmented the visibility into this trading was in real time. Between them, Susquehanna and Citadel Securities are now alleging roughly $99 million in combined losses against a scheme both peg at $100–137 million in total profit — meaning, on either firm’s number, close to the entire windfall was extracted from just two large, sophisticated options market makers, not spread across the market

.

Both figures come from each firm’s own pleadings, not a shared or independently audited count — which is itself the point: the two largest counterparties to this scheme can’t agree on its size, and both firms’ recovery is capped and offset under the same Section 20A rules regardless of whose number is closer to right.

The case has gone quiet exactly where it needed to move

As of this writing, six weeks after the complaint was filed, the public docket shows no ruling on the July 10 show-cause hearing, no unmasked defendant, and no reported outcome on Citadel’s motion to join — the last indexed filing on the public docket is the June 29 temporary restraining order itself (case docket). That’s worth sitting with, because the July 10 hearing was the case’s first real test of whether the freeze order would produce an identified defendant or simply get contested and unwound. Silence here isn’t necessarily bad news — sealed filings, ongoing subpoena compliance, or a docket-mirroring lag are all plausible — but it means the single most consequential open question in this case (does anyone actually get named, and does anything actually get frozen for keeps) remains unanswered in public exactly at the moment the story needs it to move for either firm to collect anything.

A second, separate legal front opened in the interim. On July 13, a securities-fraud class action against Futu Holdings itself — filed on behalf of Futu shareholders, not the market makers, via Glancy Prongay & Murray LLP — alleges Futu and its CEO made materially false or misleading statements by failing to disclose that the company was operating an unlicensed mainland-China securities business likely to draw penalties, covering a class period from May 24, 2023 through May 27, 2026, with a lead-plaintiff deadline of August 25, 2026 (GlobeNewswire). This is a legally distinct claim from Susquehanna and Citadel’s — a Rule 10b-5 issuer-disclosure suit brought by Futu’s own shareholders, not the Section 20A contemporaneous-trader suit brought by the market makers against the alleged inside traders — and the two should not be conflated. But it confirms the underlying fact pattern is now attracting legal scrutiny from a second, independent direction: the same regulatory-dialogue timeline Susquehanna’s complaint says leaked to options traders is also, separately, now the basis for a claim that Futu itself misled the market about its exposure to it.

The consensus read, stated plainly

Every account of this story so far — the wire reports, the aggregators, the trade press — converges on the same frame: unidentified, sophisticated traders exploited leaked Chinese regulatory information for an extraordinary payoff; Susquehanna is suing to unmask them and recover its loss; the SEC and DOJ are now investigating; justice, in some form, is likely to follow. That’s a reasonable summary of the public record, and nothing in it is wrong. It’s also not the part of the story that matters most to anyone who prices risk in these names, because it treats the lawsuit as a recovery mechanism when the statute it’s built on was written to make sure it mostly isn’t one — and it treats the surveillance apparatus that exists around U.S. options markets as more real-time than it actually is.

Section 20A was not built to make Susquehanna whole

Susquehanna’s first claim is brought under Section 20A of the Securities Exchange Act, 15 U.S.C. § 78t-1, the statute that gives a “contemporaneous trader” — someone who bought or sold the same security at the same time as an alleged inside trader — a private right of action. It’s the correct and really the only available vehicle for a market maker in this position; Section 20A exists specifically because ordinary Rule 10b-5 fraud claims are hard for a market-neutral counterparty to plead. But the statute has two built-in limits that the coverage of this case has mostly skipped past.

First, subsection (b)(1): damages “shall not exceed the profit gained or loss avoided in the transaction or transactions that are the subject of the violation.” Susquehanna isn’t suing for punitive damages or for some multiple of its loss — its own pleading caps its ask at $71.4 million, which is the ceiling the statute allows regardless of how bad the conduct was (15 U.S.C. § 78t-1).

Second, and more consequential, subsection (b)(2): “the total amount of damages imposed against any person under subsection (a) shall be diminished by the amounts, if any, that such person may be required to disgorge, pursuant to a court order obtained at the instance of the Commission... relating to the same transaction or transactions” (15 U.S.C. § 78t-1). In plain terms: Susquehanna’s private recovery and the SEC’s public enforcement recovery are not additive. They draw from the same pool. If the SEC obtains a disgorgement order against a given defendant for the same trades, Susquehanna’s own judgment against that defendant shrinks by exactly that amount. The two-track process running in parallel right now — Susquehanna’s civil suit alongside the SEC’s and DOJ’s own investigations — looks to an outside observer like redundancy, extra pressure on the defendants, more paths to justice. Legally, it’s closer to a queue. Whichever process actually collects money from a given defendant first determines what, if anything, is left for the other to collect. Since the SEC’s parallel civil enforcement process and DOJ’s criminal process have priority under the statute’s own language — notwithstanding that it was Susquehanna’s private suit, not either regulator, that produced the case’s only freeze order so far — Susquehanna’s civil case is still functionally subordinate to outcomes it doesn’t control once a regulator moves.

Want the dollar-for-dollar version of this math? In a deeper research note for paid subscribers, I run Susquehanna’s own $71.4 million loss all the way through the statutory cap, the disgorgement offset, and the cross-border enforceability discount — the same three-step chain, with real numbers at each step: Section 20A caps and offsets: the full recovery math, on Patreon.

This is why the subpoena order Susquehanna won on June 30 is the real substance of the case, not a procedural footnote. Section 20A liability requires knowing the defendant’s identity to enter judgment against; before that, discovery — subpoenaing Interactive Brokers, TradeUP, and Futu’s platform for account records — is the only thing a “John Does 1-100” complaint can actually produce. The lawsuit’s realized value in the near term is identification, which has genuine worth: it’s the mechanism that turns anonymous accounts into named defendants the SEC and DOJ can act against, and it puts confidential trading records into a public docket that deters the next attempt. But identification is not the same claim as recovery, and the statute Susquehanna sued under makes recovery contingent on facts — asset location, and who gets there first — that remain unknown seven weeks into the case. The nearest test of that was the court’s July 10 show-cause hearing, where any account holder wanting the freeze on their funds lifted had to appear and post a $100,000 bond to do it — the first mechanism, short of a successful subpoena, that could put a name to any of the John Does (Philadelphia Inquirer). As noted above, no outcome from that hearing has surfaced in the public docket as of this writing.

Citadel’s own filing is a live demonstration of the offset mechanism

Section 20A’s offset language, above, might read as a technical wrinkle until you see it playing out in real time. Citadel Securities’ July 2 motion to join Susquehanna’s suit states its rationale plainly: it is seeking party status “partly because any recovery by Susquehanna, including from the frozen funds, could reduce the amount available to compensate its own losses” (Claims Journal/Bloomberg). That is not a hypothetical about SEC disgorgement diminishing a private judgment — it is one contemporaneous trader telling a federal court, in writing, that another contemporaneous trader’s success in the same case is a threat to its own recovery, because both are drawing against the same frozen accounts and the same pool of defendant assets. This is the clearest evidence available that the “shared pool, not additive recovery” reading of the statute above isn’t just a theoretical interpretation — the two largest alleged victims are already litigating on that premise before a single defendant has been named.

It also reframes what “more plaintiffs” means here. Two large market makers pursuing the same John Does looks, from a distance, like additional pressure that should improve enforcement odds. Legally, within the frozen-asset pool specifically, it is closer to two creditors filing claims against the same limited estate: it does not enlarge what is collectible, it determines how a fixed and currently unknown amount gets divided if and when it is collected. Susquehanna’s $71.4 million ask and Citadel’s roughly $28 million claim, filed against the same frozen accounts, avoid competing with each other only to the extent the identified defendants turn out to hold assets well in excess of $99 million combined — a strong assumption for defendants the complaint itself argues are most plausibly China- or Hong Kong-based individuals, not a well-capitalized institution.

Why the money likely doesn’t come back even after identification

Assume Susquehanna’s subpoenas work and the accounts get unmasked. The next question is whether a U.S. court judgment against those individuals is worth anything, and that depends entirely on where they and their assets are. The complaint’s own theory of the case narrows this considerably: it argues the two plausible sources of the leak are Chinese securities regulators — CSRC staff, whom the complaint says are barred from trading on inside information under Article 51 of China’s Securities Law, citing the National People’s Congress’s own official English translation of the statute — or personnel at Futu or UP with knowledge of the CSRC’s pre-announcement dialogue with the companies (complaint). Both categories point toward individuals based in mainland China or Hong Kong, not the United States. That inference is reinforced by the trading pattern itself: roughly three-quarters of the flagged volume moved through Interactive Brokers, TradeUP, and Futu’s own platform — and TradeUP and Futu’s platform are subsidiaries of the very companies at the center of the leak, used heavily by mainland Chinese retail and institutional clients trading U.S.-listed names.

A U.S. federal court has personal jurisdiction here because the defendants traded on U.S. exchanges through a U.S.-based brokerage account — that’s the long-arm-statute theory the complaint pleads, and it’s a solid one for getting the case into court (complaint). It says nothing about whether a judgment can be enforced once entered. There is no bilateral treaty between the United States and China on recognition and enforcement of civil judgments, and China is not a signatory to any multilateral convention that would fill that gap (Harris Sliwoski, China Law Blog). Mainland Chinese courts can recognize a foreign judgment only under a reciprocity principle that isn’t codified in a way plaintiffs can rely on in advance, and there is exactly one documented instance of it working in China’s favor for a U.S. judgment: in 2017, the Wuhan Intermediate People’s Court recognized and enforced a Los Angeles Superior Court default judgment in Liu Li v. Tao Li and Tong Wu, reasoning that reciprocity had already been established by an earlier U.S. federal court’s enforcement of a Chinese judgment (CMS Law). That single, non-binding case is the strongest precedent available; it is not a rule a plaintiff can count on. Hong Kong is procedurally friendlier but not categorically different — the United States isn’t among the roughly 15 countries covered by Hong Kong’s statutory reciprocal-enforcement ordinance, so a plaintiff has to fall back to the common-law route: filing a fresh writ and proving the U.S. judgment is final, conclusive, and a fixed money award against a specific person, a real but slower and costlier path than the ordinance covers for judgments from, say, Australia or Singapore (Timothy Loh LLP).

None of this means recovery is impossible. If a defendant turns out to hold U.S.-reachable assets — a brokerage account itself can sometimes be frozen and attached before funds move — collection is straightforward. Susquehanna’s freeze/subpoena strategy targets exactly that possibility, and it’s the right first move for that reason. But the base case, given where the complaint itself says the informational edge originated, is defendants and assets outside the reach of a U.S. judgment. Section 20A gives Susquehanna a five-year statute of limitations to keep trying — the clock runs from the last transaction at issue, so into 2031 — which is long enough for asset locations to change, for related criminal proceedings to develop, or for a settlement to emerge under pressure that a judgment alone wouldn’t produce. That’s a real avenue. It’s a slower and less certain one than “the SEC and DOJ are on it” implies.

This isn’t a one-off tail event — it’s a structural gap with a documented precedent

The consensus framing treats this as an aberration: a uniquely brazen scheme that happened to get caught. The available evidence suggests something closer to a predictable outcome of a structural mismatch between how long Chinese insiders typically sit on price-relevant information and how U.S. market surveillance is calibrated to catch informed trading.

A study by Pengfei Ye and coauthors at Virginia Tech and the Shanghai University of Finance and Economics, published in the Journal of Accounting and Economics, examined China’s 2017 “sell-by-plan” mandate, which requires corporate insiders to pre-disclose stock sales 15 trading days in advance specifically to prevent them from dumping shares just ahead of bad news. The study found that Chinese insiders typically learn about negative news at least 25 trading days before it becomes public — ten days longer than the mandate’s waiting period was designed to neutralize, meaning insiders can schedule a “compliant” sale that still front-runs the information gap (Virginia Tech News). That’s a finding about corporate executives front-running their own company’s bad news inside a specific equity-sale disclosure regime — a different kind of insider (a company executive) and a different kind of information (the firm’s own undisclosed financial condition) than a regulator’s staff or company employees allegedly leaking a third party’s enforcement timeline to outside options traders, which is Susquehanna’s theory here. It shouldn’t be read as direct evidence about who traded FUTU and TIGR puts. What it does establish is a base rate: in the Chinese regulatory and corporate environment specifically, price-relevant information routinely leaks weeks before public disclosure, not days. Susquehanna’s own two-week trading window fits comfortably inside that documented lead time.

U.S. surveillance infrastructure is not calibrated for a 25-trading-day leak window. FINRA’s Insider Trading Detection Program monitors “100% of trading in stocks, options and bonds” using the Consolidated Audit Trail and describes itself as tracking activity “around material news events” — but by FINRA’s own account, in a 2024 podcast describing its process, a single investigation “can take anywhere from six to eight months” before it produces a referral to the SEC, and referrals, not trading halts, are the output (FINRA). That’s a detection-and-prosecution pipeline running after the news event, not a pre-trade filter. In the most recent year FINRA has publicly disclosed a figure for (2023, per that same 2024 disclosure), it generated more than 450 referrals — real evidence the system works as designed — but “as designed” means it identifies suspicious activity for later enforcement, not that it stops a market maker from being the counterparty in real time. Susquehanna was still selling puts to these accounts on May 21, the day before the news broke, despite 200,000-plus contracts of unusually concentrated flow already having accumulated over the prior two weeks. Nothing in FINRA’s own description of its process suggests that outcome is a failure of the system; it’s what the system, structured as a post-hoc referral engine, is supposed to produce.

Layered on top of that timing gap is a visibility gap the SEC has flagged in a different but related context. In a staff bulletin on foreign omnibus accounts, the SEC warns that a network of foreign financial institutions and U.S. broker-dealers can be structured so that “none of them has complete visibility into the total amount of securities deposited, trading volume and activity” behind an account, because the ultimate beneficial owner of the funds is “unknown to a broker-dealer because of the omnibus account structure” (SEC staff bulletin). That bulletin addresses low-priced-securities fraud schemes specifically, not this case — it is not evidence about FUTU or TIGR — but the structural point it describes transfers directly: a large share of the Subject Trades here ran through TradeUP and Futu’s own platform, broker-dealers owned by the companies whose confidential regulatory information was the alleged source of the leak. That is close to a worst-case configuration for a surveillance system built around cross-broker pattern detection: the entity best positioned to know when an insider is trading through its own platform is the same entity whose employees are named as one of the two plausible sources of the tip.

Why the trade was built loud, not quiet — and what that bounds

A reasonable question about the mechanics: why would traders sitting on privileged information precise enough to target expirations inside a week of an unannounced date trade in a way almost guaranteed to draw attention, rather than camouflaging the position? The academic literature on informed trading gives a direct answer, and it doubles as the decay and capacity analysis this kind of claim requires.

Kyle’s foundational model of informed trading, and the stealth-trading literature built on it, predicts informed traders split large positions into smaller trades over time specifically to avoid revealing their information before they can profit from it. A recent formal treatment, “Wealth or Stealth? The Camouflage Effect in Insider Trading” (Ma, Xia, and Zhang), models this as a trade-off governed by legal risk: insiders trading among a population of ordinary liquidity traders can camouflage their activity, but doing so caps how much of their informational edge they can extract, because size and speed increase both price impact and detection probability. The paper’s equilibrium result centers on what it calls a stealth index — a parameter pinning down how much an insider population trades, given the balance between expected profit and prosecution risk.

The Subject Trades in Susquehanna’s complaint sit at the wealth end of that spectrum, not the stealth end. Eighty-seven to 98% concentration in specific contracts on specific days, nine accounts responsible for all the relevant Interactive Brokers flow, two-thirds of the position expiring within a week of an unannounced date — none of that is camouflage; by the standard the literature uses to define stealth trading, it is close to the opposite. That is informative about the traders’ own assessment of their position, not just their recklessness. A trader confident their information window was narrow and closing fast has a rational reason to prioritize size and speed over stealth: spreading the same trades over eight weeks instead of two would have meant less price impact per trade, but it would also have required a longer information lead time than even the roughly 25-trading-day Chinese-insider base rate reliably supports, and it would have left more time for the position to leak or for the stock to drift on unrelated news. The trade’s loudness is evidence of urgency, and urgency is evidence the information had a short, specific shelf life — a stronger and more precise inference than “some buyers had material non-public information,” and one the complaint’s own strike and expiry data supports without needing the traders’ identities at all.

That mechanism sets the decay and capacity bounds directly. Decay of this specific tactic: a wealth-style, concentrated, affiliate-broker-heavy signature is now a documented pattern — named in a federal complaint, covered by Bloomberg and Reuters, and actively probed by the SEC, DOJ, and two of the largest options market makers in the country. Anyone attempting the identical playbook in a comparable name faces materially higher detection odds than the traders here did in May, precisely because this case is now the reference pattern surveillance teams and market-maker risk desks will match against; that specific signature should decay fast. Decay of the underlying edge, by contrast, should not: the source of the informational advantage — long Chinese-insider lead times on regulatory bad news, and weak U.S. enforceability against China- and Hong Kong-based defendants — is a structural feature of the jurisdictional and disclosure environment, not an artifact of one scheme, and neither the Ye et al. lead-time finding nor the treaty gap changes because a single case got prosecuted. Capacity: the fact that essentially the entire alleged profit pool — $100 million to $137 million — was extracted from just two large options market makers rather than diffused across dozens of smaller counterparties indicates the ceiling on this kind of trade is set by how much short-dated, deep-OTM put risk a small number of major liquidity providers are willing to warehouse in mid-cap, single-name China ADRs around any given expiration. Push the same trade to a materially larger size in the same names and window, and the buying pressure itself would be expected to move implied volatility enough to signal the position before expiration — the trade partially self-detects at scale, the same constraint the camouflage model formalizes. The edge is real, structural, and durable at the informational-asymmetry level; it is not scalable at the execution level much beyond what already happened here without triggering the visibility that produced this lawsuit.

That has a direct implication for who actually bears the risk. It is not the market broadly — it is specifically the handful of firms with enough capital and risk appetite to be the marginal seller of size in short-dated OTM puts on names like FUTU and TIGR. A market maker that is not among the two or three dominant liquidity providers in a given China ADR’s options chain is not exposed to this version of the risk regardless of its overall book size; this is a name-selection and flow-monitoring problem, not an AUM-scaling one.

The obvious objection

A sophisticated reader’s response to all this is predictable: the SEC and DOJ are already investigating, FINRA covers 100% of the market, and this is exactly the kind of egregious, coordinated pattern that regulatory infrastructure is built to catch and punish — so isn’t the pessimism about recovery overstated? Doesn’t “the system worked, just slower than we’d like” undercut the structural argument?

It doesn’t, because detection and recovery are different outputs measured on different clocks, and the market maker’s loss is realized on neither of them. The loss crystallizes the moment Susquehanna sells the put and the underlying gaps down — that’s instantaneous, on May 22. FINRA’s own six-to-eight-month investigation timeline, followed by whatever additional time an SEC civil case or DOJ criminal case takes to reach judgment, followed by the enforceability questions above, means the earliest plausible point at which money could theoretically flow back to Susquehanna is measured in years, not weeks, and is conditional on facts — asset location chief among them — that remain unknown, and that the docket’s silence since June 29 has done nothing to resolve. A system can work exactly as designed, generate a referral, produce an indictment, and still deliver a market maker zero dollars of recovered loss, because “working as designed” was never a promise about restitution timing or collectability. The objection conflates enforcement succeeding with the plaintiff getting paid; Section 20A’s own offset language is proof Congress understood those as separate outcomes when it wrote the statute in 1988.

A second version of the objection: doesn’t Citadel joining strengthen the case — more capital, more pressure on the eventual defendants? For deterrence and discovery, yes. For recovery, no: Citadel’s own stated motive for joining is defensive, not additive, which only makes sense if both firms already expect the collectible total to be the binding constraint. Two plaintiffs against the same frozen accounts split a fixed recovery; they don’t multiply it.

📊 Want Deeper Quantitative Analysis?

This research took a very long time of data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

Follow more of my work on YouTube and LinkedIn.

Optiver’s June 2026 paper on European market structure reads like neutral policy analysis: five technocratic fixes to stop the decline of lit continuous trading, each grounded in ESMA’s own data. Read against who benefits, the same five fixes form a coherent competitive-positioning document for systematic internalisers — the exact business Optiver itself runs. That doesn’t make the underlying facts wrong. It does mean the facts need to be checked against primary sources rather than taken as neutral, and at least one of them — the retail-venue evidence — is a 2022 study of a practice that stayed partly legal in the EU until 21 days after this paper was published, a timing gap the paper doesn’t flag.

The paper and the trend it describes

On June 9, 2026, Optiver — a Dutch market maker and, since 2025, a licensed systematic internaliser (”SI,” meaning a firm that fills client orders using its own capital and balance sheet rather than matching two outside counterparties) — published Building stronger European equity markets, a five-point reform package aimed at the EU’s Market Integration and Supervision Package (MISP), the Commission’s package of proposals published December 4, 2025. ECON’s three rapporteurs published draft amendment reports on June 12, 2026; other committee members may table further amendments by July 16, 2026, and ECON intends to vote on its negotiating position at its December 1, 2026 meeting — the actual near-term legislative timeline, which is a committee vote on a negotiating mandate, not a final agreement. The paper’s factual foundation is ESMA’s April 30, 2026 Call for Evidence on the market structure of European equity markets, a data-driven study of 2022–2025 MiFIR transaction reporting that both sides of the debate cite as ground truth.

ESMA’s own numbers confirm the trend: lit continuous trading — order books where displayed buy and sell quotes trade against each other continuously, setting the reference price everyone else free-rides on — fell from an average 84% to 78% of transaction count between 2022 and 2025, while overall on-book trading (continuous books plus auctions) held at “around 75-80%” of turnover and addressable liquidity (trading genuinely open to any counterparty, not internal or technical bookings) stayed roughly stable at 85-90%, per ESMA’s Call for Evidence, paragraphs 21 and 38. Optiver’s quarterly figures for the two channels absorbing the shift both check out: non-intragroup SI trading grew “from 5.1% in Q1 2022 to 10% in Q4 2025,” a verbatim match to ESMA’s paragraph 34, and frequent batch auctions (FBAs, ultra-short automated auctions lasting milliseconds) reached “7.9% in Q4 2025,” matching ESMA’s paragraph on FBAs doubling their turnover share in three years. ESMA’s prose gives period averages; Optiver’s gives quarter-end figures, which run higher if the trend is still accelerating into Q4 2025 — a measurement-convention difference, not a contradiction.

ESMA’s own report frames this as a live dispute, not a settled question. It names two competing outside studies directly: Oliver Wyman’s “The Liquidity Matrix,” commissioned by the Federation of European Securities Exchanges (FESE) in July 2025, which frames intra-market fragmentation as a threat to price discovery and capital-raising; and Goldman Sachs and New Financial’s October 2025 report, which argues that focusing on the lit primary market captures barely 30% of real trading activity and that European markets are, in the round, working well. This is the consensus a reader needs before assessing Optiver’s paper: the exchange lobby (FESE) says fragmentation is a crisis; a bank that itself supplied over €1 trillion of principal liquidity in 2024 says the crisis is a mirage. Both are talking their own book. Optiver’s paper does something more interesting than either — it accepts the lit-decline premise (siding with the exchanges’ diagnosis) while locating the blame and the fix somewhere that benefits neither the exchanges nor the banks, but market makers like itself.

Reading the five proposals by who gains

Optiver’s own paper discloses, in its second section, exactly what kind of firm is writing it: “at Optiver, when we transact via our SI, the resulting risk is managed passively... We believe this model... is how a well-functioning risk-taking SI should operate.” That sentence is the interpretive key to the rest of the document. Run each of the five proposals through it:

Proposal 1 (midpoint pricing for lit venues) asks regulators to extend to exchanges a permission SIs already have. Optiver is an SI. If exchanges get midpoint flexibility, Optiver’s SI stops holding an exclusive advantage over exchanges on this dimension — but nothing here would remove the SI’s advantage over other SIs or over single-market-maker venues (proposal 5’s target). This is a genuine give-back to competitors, not a pure self-benefit, and it is the one proposal where Optiver’s paper is arguing against its own narrowest interest.

Proposal 3 (compulsory intragroup flagging) would make post-trade data cleaner for everyone. But it specifically benefits independent SIs like Optiver relative to bank-affiliated SIs and exchange groups that route large volumes of internal, group-affiliated flow — the flagging gap Optiver describes inflates the appearance of competitive bilateral liquidity that is actually captive. ESMA’s own Call for Evidence, in its section on addressable liquidity (paragraphs 122-123), independently arrives at the same conclusion — intragroup SI trades “do not seem to qualify as addressable liquidity” — and asks stakeholders directly, in Q44, whether a new RTS 1 flag should be created. Optiver’s ask here tracks ESMA’s own emerging thinking; it is not solely self-serving, but it does help precisely the kind of firm Optiver is.

Proposal 4 (full clearing interoperability) would let any trading firm pick its own central counterparty (CCP — the institution that steps between buyer and seller and guarantees settlement) rather than defaulting to the CCP owned by the exchange group it trades on. Optiver benefits directly: full interoperability removes the fixed cost of maintaining separate clearing relationships per venue and lets it compete for flow at venues currently locked to a vertically integrated incumbent. Optiver isn’t alone in pushing this — a July 2026 joint letter from the European Banking Federation, AFME, Cboe, EFAMA, and EPTA (EBF’s release lists the CCP signatory simply as “Cboe”; it is Cboe Clear Europe) calls for the identical mandate inside MISP. Every signatory has a direct commercial stake: Cboe Clear Europe is a rival CCP that stands to gain market share, and it already reports 95% access to European equities trading but only 73% of that flow under full interoperability today, 22% still under the weaker “preferred” model. EPTA represents exactly the class of firm — independent principal traders — that gains most from breaking vertically integrated clearing monopolies.

Proposal 2 (consolidated-tape venue attribution and depth) asks that the forthcoming EuroCTP consolidated tape attribute trades to venues with more granular depth-of-book detail than a flat, top-of-book-only feed. This is a smaller and more technical beneficiary claim than proposals 1, 3, or 4, but it still has a beneficiary: as one of a handful of large SIs that already account for a concentrated share of non-lit turnover (see below), Optiver gains when its own liquidity contribution is visible and creditable in the consolidated record in a way a flatter, less granular tape would not capture — a real interest, though a modest one relative to the other four asks.

Proposal 3, continued (MMT as the compulsory flagging standard): Optiver’s paper states “EPTA has called for compulsory MMT adoption” as supporting evidence. EPTA’s own membership page, published by the FIA, lists Optiver VOF as a current member alongside Citadel Securities, Jane Street, Susquehanna, Virtu, and fifteen other principal trading firms (twenty members in total). Optiver is citing its own trade association’s position as third-party corroboration without disclosing the membership — and the tie is closer than membership alone: Lotte de Vos, Optiver’s Head of European Market Structure, was elected to FIA EPTA’s eight-member voting Executive Committee on June 16, 2025, the body that sets EPTA’s policy positions, including the MMT position cited here. This is not evidence the underlying MMT proposal is wrong — voluntary flags genuinely have had weak uptake, a fact independent of who is asking for a fix — but it is evidence that “the industry agrees” and “Optiver agrees” are, in this instance, closer to the same sentence than the paper lets on.

Proposal 5 (ban single-firm exclusivity on retail venues) is the one built on the shakiest and most dated evidentiary base, addressed on its own below, because it is also the proposal where Optiver has the least direct commercial stake and the strongest case — which makes the sourcing problem worth fixing rather than dismissing.

The retail-venue evidence is real, and it is four years stale

Optiver’s paper supports proposal 5 with two regulator studies: “the AFM found that at two of these so-called single market-maker venues... between 72% and 83% of retail orders were executed at worse prices than the reference price, at an average cost of up to 11 basis points per trade. The CNMV found that 86% of trades at a comparable venue were executed outside the contemporaneous price range of the ten most liquid competitive venues.”

Both figures check out against the primary documents. The AFM’s March 2022 paper reports “PFOF trading venue X” at 68.8-72.0% worse executions (4.8bps average deterioration) and “PFOF trading venue Y” at 81.5-83.3% worse (11.5bps) — both described as having “one market maker acting as the counterparty for nearly all retail client orders in shares.” Optiver’s “72% and 83%” is the pair measured against Euronext Amsterdam specifically, and “up to 11 basis points” rounds down slightly from 11.5bps — a minor, immaterial imprecision. The CNMV’s March 2022 companion study reports 86.4% of trades as “worse execution,” an average deterioration of €1.09 per €1,000 — Optiver’s “86%... approximately €1 per €1,000” matches that directly.

What the paper doesn’t say: both studies measure 2021 data, published February-March 2022, and every venue in both is explicitly a “PFOF trading venue” — the broker was paid by the market maker to route retail orders there. That practice isn’t a live policy question anymore. Article 39a of the revised MiFIR bans an investment firm from receiving “any fee, commission or non-monetary benefit from any third party” for routing retail or opt-in-professional orders to a venue, permitting a transitional exemption only “until 30 June 2026.” Germany was the sole member state using it; once that window closed, PFOF became illegal EU-wide — June 30, 2026, 21 days after Optiver published this paper. The arrangement that generated the AFM and CNMV’s worst-execution numbers is now illegal across the bloc, though the ban post-dates rather than precedes Optiver’s analysis.

This is where the paper’s argument needs to be read carefully rather than dismissed. Optiver is not asking MISP to ban PFOF — that has already happened. It is asking MISP to address something distinct: venue rulebooks that grant one firm exclusive liquidity-provision rights regardless of whether a payment changes hands, “in some cases... written into venue rules,” which the paper correctly notes makes a venue “legally classified as open, multilateral” while “operat[ing] like a bilateral venue in practice.” The AFM and CNMV studies measured venues that happened to combine both features — single-dealer structure and PFOF payment — because in 2021 those two things nearly always went together. The ban removes the payment. It does not, on its own, remove the exclusivity clause in the rulebook, and neither the AFM nor the CNMV has published updated execution-quality data isolating the structural effect (single dealer) from the payment effect (PFOF) after the ban. Optiver’s proposal 5 is a genuine, falsifiable empirical claim — that a single-dealer venue will keep producing worse retail prices even with no payment attached — resting on quantitative evidence that predates the one intervention (the PFOF ban) a skeptical regulator would point to as already having solved the problem. The paper would be measurably stronger, and more honest about its own limits, if it said so.

This is also not a new argument Optiver assembled for the MISP moment. Optiver made substantially the same point in its own March 24, 2025 paper — over a year before the June 2026 paper, and before the PFOF ban had even taken effect — warning that “ahead of the upcoming PFOF ban, some firms are introducing new structures that directly link single market-maker venues with affiliated brokers,” concluding that these post-ban structures “offer even less competition for retail order flow than their predecessors” because the affiliated market maker “no longer even has to pay for retail order flow, but may still get to execute it on highly preferential or exclusive terms.” That earlier paper cites the same AFM and CNMV studies used in the June 2026 paper, which undercuts any reading of proposal 5 as manufactured for this specific legislative window, though it does not supply a newer quantitative dataset either.

Separately, live 2026 evidence — not from Optiver, outside either regulator’s 2021-22 sample — corroborates the structural mechanism, if not the execution-quality numbers. Scalable Capital’s own European Investor Exchange (EIX), a post-ban venue the broker co-founded, “technically has two market makers,” but its own rulebook states: “the Management assigns each security to one of the authorized market makers for quoting purposes” — one per instrument in practice. The independent analysis documenting this argues the resulting “closed-loop” structure, where the broker now keeps the full spread rather than a partial PFOF kickback, may leave the conflict of interest worse, not better, after the ban — real, current, named confirmation that the structure proposal 5 targets exists post-ban, though only a qualitative, one-venue illustration, not a replacement for updated execution-quality data, which still doesn’t exist for any post-ban venue.

The cap-failure history is accurate, and the cap regime has already moved on

Optiver’s paper leans on the double volume cap (DVC) — MiFID II’s original mechanism for limiting dark trading under the reference-price waiver, first applied in March 2018 — as its proof that “caps have been tried, they did not work”: each time the cap suspended dark trading in a stock, activity reappeared elsewhere, mostly in frequent batch auctions, rather than returning to lit continuous books. That history is accurate. What the paper does not mention is that the DVC no longer exists in the form it describes: the 2024 MiFIR review replaced it with a single volume cap (SVC), effective October 2025 — one EU-wide 7% ceiling, replacing the old two-tier per-venue-and-EU-wide system, with the same three-month suspension if breached. ESMA itself says it “will continue to monitor the developments in the market... from a double volume cap to a single volume cap.”

That matters for weighing the “caps don’t work” argument: it’s a historical pattern observed under a specific, now-retired mechanism, not a timeless law, and the SVC’s single EU-wide threshold is a genuinely different design from the old dual cap that produced more numerous, more gameable suspension events. Whether the same migration-to-FBA pattern repeats under the SVC is an open question ESMA is actively collecting data on, not a settled precedent. Using the DVC’s 2018-2023 record to forecast the SVC’s behavior from October 2025 onward extrapolates across a regime change the paper doesn’t flag.

Why the SI-share edge exists, and where it runs into a ceiling

None of the above explains why non-intragroup SI trading could grow the way it did — from 5.1% to 10.0% of turnover in three years, doubling — or why that growth is likely to run into a natural limit rather than continuing indefinitely. Both questions matter more to a trading desk than the lobbying story, because they determine whether “more flow moves to SIs” is a trend worth positioning around or one already close to exhausted.

The mechanism is two-sided. On the regulatory side, ESMA’s own Call for Evidence, paragraph 77 states the 2024 MiFIR review’s SI-only midpoint permission “might have redirected certain trading flows from trading venues to SIs” — a real, ESMA-acknowledged regulatory subsidy toward the SI channel that has nothing to do with SIs offering better fundamental execution and everything to do with a tick-size rule applying unevenly. On the demand side, institutions with orders larger than the size displayed at the best price genuinely need a counterparty willing to commit capital at one price rather than walking the book — the second form of “invisible” price improvement Optiver’s paper describes, where an SI fills 5,000 shares at the best price instead of forcing the investor to pay up through five price levels to complete the order.

Both of those growth drivers point to their own limits. The regulatory driver is the more fragile one: it should erode substantially, though probably not disappear entirely, the moment proposal 1 succeeds and lit venues get the same midpoint permission — meaning a mechanism that has been feeding Optiver’s SI volume for two years is one Optiver’s own paper is asking regulators to switch off. It would not disappear completely because other, price-independent reasons to route to an SI would remain even after midpoint parity: established counterparty and credit relationships, KYC/onboarding requirements that make switching venues costly, and the information-leakage protection a bilateral SI fill offers relative to displaying an order on a lit book. That is still a genuine, checkable prediction: if MISP grants venues midpoint parity, non-intragroup SI growth should flatten or partially reverse in the following data — not necessarily vanish outright — decoupled from any change in genuine investor demand for committed-capital execution.

The demand-side driver is bounded differently — by concentration and by price-discovery mechanics academic research has already mapped. ESMA’s own SI data shows the growth is not broad-based: “five (ten) SIs concentrate 50% (80%) of turnover and 59% (83%) of number of trades in 2025,” per the Call for Evidence’s section 4.3 — “SI trading is growing” is largely a statement about how much balance sheet a handful of large market makers will commit, not a diffuse shift, and that capacity is itself constrained by how much single-name risk those firms will warehouse before their own risk limits bind. A body of market-microstructure research converges on the same structural conclusion. Zhu’s model, an NBER working paper (2011 draft, published 2014 in the Review of Financial Studies) finds that “adding a dark pool tends to concentrate payoff-relevant information onto the exchange and, under natural conditions, improves price discovery” — but only under those conditions. Ye (2016), extending Zhu’s framework, shows the sorting logic reverses when information is noisy: “when information precision is low... the majority of informed traders... prefer a dark pool... impairing price discovery in the exchange.” And Foley and Putniņš’s natural-experiment study of Canada and Australia’s dark-trading restrictions, published in the Journal of Financial Economics (2016), finds dark limit order markets generally help market quality, while dark midpoint-crossing systems help only conditionally — a CMCRC summary describes the midpoint-crossing finding as “the existence of a ‘tipping point’ beyond which dark trading is detrimental,” language not in the paper’s own abstract but consistent with its design and its authors’ framing that the tipping point “is likely to differ between markets.” Together these support a directional, not numerical, conclusion: off-lit execution depends on lit venues producing a reliable reference price, so migration away from lit trading can’t run to 100% without undermining the signal the SI and auction channels price off. No source here gives a universal percentage threshold, and the precise European tipping point isn’t independently observable from public data — but the direction holds: this is a self-limiting mechanism, not “SIs are simply better and will keep taking share.”

The obvious objection, and why it only partly holds

A sophisticated reader’s response to all of the above is: so what if Optiver benefits — does that make the analysis wrong? No, and it shouldn’t be read that way. Proposals 1 through 4 rest on facts that are independently verifiable against ESMA’s own primary data, not against Optiver’s say-so: the midpoint asymmetry between SIs and venues is confirmed in ESMA’s Call for Evidence, paragraph 77, which states in its own words that the 2024 MiFIR review’s SI-only midpoint permission “might have redirected certain trading flows from trading venues to SIs” — language Optiver quotes accurately. The RPRI price-improvement blind spot is confirmed in the same document (paragraph 91): RPRI-flagged trades “account for only a negligible portion of total turnover,” which ESMA itself flags as possibly reflecting “a data quality issue” rather than an absence of real price improvement — the exact point Optiver’s paper makes about midpoint fills not registering as “price improvement” under a narrow flag definition. Self-interest explains why Optiver chose to write about these particular facts. It does not, on the evidence gathered here, mean the facts are misstated.

But “the facts are accurate” and “the proposal is good policy” are different questions. On proposal 1: the mechanism evidence gathered above (Zhu, Ye, Foley & Putniņš) shows off-lit trading is self-limiting, but it doesn’t independently establish that midpoint parity for lit venues specifically improves outcomes — so proposal 1 stands or falls on the ESMA data (the midpoint asymmetry is real), not on a settled efficiency case. On proposal 3 (intragroup flagging): ESMA’s own Call for Evidence independently reaches the same conclusion Optiver does — intragroup trades inflate the appearance of addressable liquidity — real corroboration from a party with no stake in the outcome. On proposal 4 (clearing interoperability): the merits case is that it cuts concentration risk in a single vertically integrated CCP and lowers fixed costs for multi-venue firms, an argument that holds regardless of Cboe Clear Europe’s own stake in it — though no independent, non-industry study is cited here beyond the coalition’s own letter. Proposals 1 and 4 survive the self-interest objection on partial evidence; proposal 3 survives it more cleanly, since ESMA’s independent finding does the confirming work Optiver’s paper cannot do for itself.

Where self-interest does matter is in what the paper omits or leaves unstated: that Optiver holds an SI license and stands to gain from proposal 1’s leveling; that Optiver is a named EPTA member citing EPTA’s position as outside validation for proposal 3; that Optiver sits in the same interoperability coalition as a rival CCP with a direct market-share incentive; and that the retail-venue evidence for proposal 5 needs a “measured before the PFOF ban, structural effect not yet isolated” caveat that isn’t in the text. None of this is unusual for an industry position paper — it is unusual for a piece written in the register of neutral data analysis rather than declared lobbying, which is exactly the register Optiver’s paper adopts throughout (”this document ... does not reflect any opinion or judgement of Optiver,” per the paper’s own disclaimer).

The rapporteur’s own draft already cuts against the clean story

The clearest evidence that MISP is not simply absorbing industry’s preferred package comes from the legislative text itself — though only from one of the three parallel MISP legislative files; the other two rapporteurs’ draft positions are not assessed here. Markus Ferber’s June 12, 2026 draft report on the MISP Omnibus Regulation — the rapporteur assigned the MiFIR-facing parts of MISP, including the SI and consolidated-tape provisions Optiver’s paper addresses — proposes requiring SIs to “publish rulebooks disclosing access criteria, execution processes and fee structures,” a “mandatory minimum price improvement of one tick size over the best lit market price for SI executions, extended from retail to all orders,” and asks ESMA to assess “whether large SIs should be subject to direct supervision.” Every one of those is a new compliance burden on SIs, not a benefit, and the price-improvement proposal in particular would apply to Optiver’s own SI business a tick-size discipline closer to the one lit venues currently carry, cutting in the opposite direction from proposal 1’s ask to relax that discipline on lit venues instead. This does not resolve which reading of the paper is correct. It does show the legislative process, at this stage, is not a rubber stamp for the SI-favourable half of Optiver’s package — which is exactly the kind of test this piece’s thesis needs to survive contact with, rather than a convenient omission.

What would prove this wrong

This piece’s claim is specific enough to fail a test. It fails if: MISP’s final text adopts proposal 5 (retail-venue exclusivity) but waters down proposals 1, 3, and 4 despite their stronger coalition backing — showing the self-interested proposals were actually weaker on the merits; a post-ban AFM or CNMV follow-up finds single-market-maker venues’ execution quality converging toward multi-maker venues once PFOF payments stop, with no further intervention — falsifying the claim that the exclusivity clause, not the payment, was doing the damage; or Ferber’s SI-transparency and price-improvement proposals survive the December 1, 2026 ECON vote largely intact alongside proposals 1-4 — suggesting ECON weighs SI-favourable and SI-restrictive ideas on their merits rather than reflecting whichever lobby wrote the loudest paper. Conversely, if proposals 1, 3, and 4 pass largely as the EPTA/Optiver/Cboe coalition proposed while Ferber’s SI-restrictive burdens get stripped out and proposal 5 stalls for lack of updated evidence, that’s the strongest available confirmation that lobbying alignment, not evidentiary strength, is driving outcomes.

What to watch

Four concrete markers, all checkable on public record over the next two to three quarters. First, ECON’s July 16, 2026 deadline for further member amendments and its December 1, 2026 vote on a negotiating position — watch whether Ferber’s SI-transparency and mandatory-price-improvement language survives alongside the clearing-interoperability and consolidated-tape articles, and whether “full” interoperability survives or gets softened back toward the “preferred” model industry operates under today. Second, whether any regulator — the AFM, the CNMV, or ESMA in its own promised Q3 2026 feedback statement on the Call for Evidence — publishes post-change data: a post-PFOF-ban execution-quality follow-up (which a robust MISP process should demand before legislating on proposal 5) or evidence on whether dark and FBA volumes are migrating under the new single volume cap the way they did under the old double volume cap. Either absence is itself informative. Third, MMT v5.0’s actual adoption rate among reporting venues over the next year — Optiver’s own paper concedes the voluntary flag has seen weak uptake for years, so a Level 1 mandate is the only lever that plausibly changes that, and its presence or absence in the final text is a clean, binary test. Fourth, and most directly tradable: non-intragroup SI share of turnover in ESMA’s next transaction-reporting update, which should flatten, not merely keep compounding, if and when lit venues win midpoint parity — the cleanest test of whether the last two years of SI growth was a regulatory subsidy or a durable investor preference. A desk pricing execution venues in European names should treat continued SI-share growth after midpoint parity as evidence the demand-side driver dominates; a flattening as evidence the regulatory driver was doing most of the work.

None of this is a reason to discount Optiver’s underlying facts. It is a reason to read a market maker’s policy paper the way an analyst reads a sell-side research note on a stock the bank also makes markets in: check the numbers independently, they will very often check out, and then ask separately who benefits from the recommendation before deciding how much weight the recommendation itself deserves.

If you found this kind of source-by-source, follow-the-incentive breakdown useful, I cover market structure and quant research like this regularly. Connect on LinkedIn or subscribe on YouTube for more.

📊 Want Deeper Quantitative Analysis?

This research took a very long time of data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

In every one-factor Hull-White implementation, dr = (θ(t) − ar)dt + σdW, mean reversion and volatility get calibrated on different clocks: volatility daily, mean reversion roughly monthly. That cadence gap is real, documented independently by a bank’s own quant team and by academic researchers, and there’s a defensible reason for it. What’s far less established is what follows from it. Exactly one vendor, one currency, and one six-month window has ever shown this gap producing a measurable valuation error — and nobody, including this piece, has shown that error converts into money changing hands. What follows separates the mechanism, which holds up, from the pattern claim, which doesn’t yet.

The consensus, and the parameter underneath it

Practitioners use Hull-White because letting θ(t) absorb the shape of today’s discount curve preserves closed-form bond and swaption prices while still fitting the market exactly (Hull and White, NYU Stern working paper). The well-known limitation is that a single Brownian driver forces every point on the curve to move in lockstep, which risk-validation teams flag as the model’s central weakness (RiskSpan). That limitation is precisely why Longstaff, Santa-Clara, and Schwartz’s 2001 study of American-style swaption exercise found that, based on ISDA notional estimates, following a myopic single-factor exercise strategy rather than a multi-factor one cost swaption holders on the order of several billion dollars in aggregate (Longstaff, Santa-Clara, and Schwartz, Journal of Financial Economics, via SSRN). None of that is new to anyone trading these books. The less-examined mechanism sits inside the one-factor model itself, in how mean reversion, a, is actually estimated day to day.

The math, stated plainly

Bond prices under Hull-White are exponential-affine

and θ(t) is solved so the model reproduces the observed instantaneous forward curve exactly:

(Brigo and Mercurio’s derivation, reproduced in “Consistency of extended Nelson-Siegel curve families with the Ho-Lee and Hull and White short rate models”). B(t,T) is the function that matters here: it’s the exponential-decay term determining how a shock to the short rate propagates along the curve, which is why a — not σ — controls the shape of the model-implied volatility surface along the tenor axis, while σ governs its shape along the expiry axis (S&P Global Market Intelligence, 2023).

Two clocks, and what’s actually corroborated

Here is the part with genuine independent support. S&P Global’s research, produced by the quant team running a bank’s own CVA/XVA calibration, states that a mean-reversion update cadence of roughly once a month, against a daily update for volatility, is what they find reasonable in practice. A separate, peer-reviewed 2025 paper in Quantitative Finance reaches the same operational conclusion from a different angle: because mean reversion and volatility have overlapping effects on Black implied vols within a single expiry-tenor smile, a is typically held constant and re-calibrated only weekly or monthly, “due to the limited impact of the mean-reversion parameter, combined with the absence of market data to calibrate this parameter to,” while volatility is re-calibrated daily (van der Zwaard, Grzelak, and Oosterlee, Quantitative Finance, 2025). Two independent teams, different institutions, agreeing on the practice itself. That’s as far as the corroboration goes, and it’s worth being precise about that boundary: neither paper, nor any other source in this piece, independently confirms any of the specific numbers in the next two sections. Those come from one place.

Where the mechanism bites hardest

This cadence gap matters disproportionately for anything with early-exercise features, because of how B(t,T) works. A Bermudan swaption’s Vega is distributed across the entire term/tenor structure rather than concentrated at one point: a representative 11-non-call-1 Bermudan receiver has a 1bp Vega of 0.06% of notional, against Vegas on its co-terminal European swaptions ranging from 0.009% to 0.048% (Gatarek and Jabłecki, Mathematics, 2021). Since B(t,T) is exactly the function mean reversion controls, a stale a misprices a Bermudan more than it misprices any single vanilla instrument used to calibrate it — and that exposure is not a corner case: roughly 60% of bonds in the Bloomberg Barclays Global Aggregate Credit Index carry call provisions that get stripped and sold on as swaptions (Gatarek and Jabłecki, 2021). For non-standard amortizing or accreting Bermudans specifically, the problem compounds further: their own calibration targets are “generally no more liquid than the [instrument] itself,” so even the calibration inputs have to be inferred rather than observed (Risk.net, “Cutting edge: Option pricing — Bounding Bermudans”). This is the strongest part of the argument, and it’s a statement about exposure, not about profit.

The one data point that exists

Benchmarking a book of at-the-money swaps out to 30 years against exact market-implied CVA, one vendor’s fully calibrated Hull-White model produced a CVA root-mean-square relative error (RMSRE) “consistently of the order of 5% or less” — except in the first half of 2022, when the euro book’s error briefly breached that threshold. Freezing mean reversion at a naive constant 5%/year instead, and recalibrating only volatility, kept CVA RMSRE largely below 10% and, in that same window, actually beat the fully optimized model (S&P Global Market Intelligence, 2023). A separate 2020 study from the same research team found optimized mean reversion for EUR, JPY, and USD all crossing into negative territory that spring — a value with no coherent identity as a “speed of reversion” (Puetter and Renzitti, IHS Markit, 2020).

That is the entire empirical record. It is one currency book (EUR), one implementation, one vendor’s research team, across one six-month window. It has not been shown to hold for USD or JPY in the same period, and it has not been tested at all against the other monetary-policy inflections of the last fifteen years — the 2015 Fed liftoff, the 2018 hiking cycle, the 2020 cutting cycle. A pattern observed exactly once, by the same team whose own methodology produced it, is a hypothesis worth taking seriously — not yet a structural, recurring feature of the market.

The gap between mispriced and profitable

Every number above is a valuation-accuracy metric: how far a model’s output sat from a market-implied benchmark. None of it is a trade, a position, or realized P&L. Showing that a bank’s own optimized calibration was less accurate than a naive constant for six months establishes that a valuation gap existed inside one firm’s risk system — it does not establish that a counterparty on the other side of a trade was pricing off the worse convention, that the gap was large enough after transaction costs and bid-offer to act on, or that anyone actually captured it. Bridging “our model disagreed with theirs” to “here is how you extract money from that disagreement” requires evidence this piece does not have: a documented instance of a specific mispriced trade, a counterparty using a demonstrably stale convention, or a realized P&L outcome tied to this specific mechanism. Absent that, the honest claim is that a real, mechanism-grounded valuation discrepancy exists and has been observed once — not that it is a capacity-bound edge with a known scale.

What would actually confirm or kill this

Confirmation requires three things this piece doesn’t have: the same RMSRE reversal appearing in USD or JPY books during the same 2022 window (testing whether it’s currency-specific or general); the same pattern appearing around 2015, 2018, or 2020 inflections (testing whether it’s a recurring feature or a one-time artifact of 2022 specifically); and a second bank’s independently built CVA system showing the same result (testing whether it’s a property of the market or an artifact of one vendor’s implementation). Any of these failing to replicate would be reason to treat the 2022 finding as an isolated event rather than a mechanism-driven pattern. None of these tests has been run in the public literature as far as this piece can establish.

What to do with this today

The mechanism — mean reversion calibrated on a slower, less-disciplined clock than volatility, with disproportionate impact on Bermudan and amortizing structures — is well enough established to be worth a risk manager’s attention as a diagnostic. A fitted mean reversion that has drifted toward zero or turned negative around a policy inflection is worth treating as a signal that the model is curve-fitting through a regime shift rather than reading it, and a CVA or exposure model on a Bermudan-heavy book is worth benchmarking against a naive fixed-mean-reversion alternative at FOMC and ECB inflection points, simply because the one time this has been tested, the naive benchmark won. That is a reasonable, low-cost risk-management practice, and it is a different and smaller claim than “this is a tradeable edge” — the evidence here supports the former, not yet the latter.

📊 Want Deeper Quantitative Analysis?

This research took a significant amount of time — data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

Connect with me:

→ Subscribe on YouTube — The Mathematical Trader

→ Connect on LinkedIn

The Avellaneda-Stoikov reservation-price framework does exactly what its founding paper says it does: it trades expected profit for a reduction in inventory variance. What the downstream literature drops is the price tag and, more importantly, its shape. The 2008 paper’s own simulation tables show the inventory-skewing strategy giving up anywhere from 0.86% to 48.76% of a naive symmetric quoter’s mean profit as risk aversion rises across the three values tested, and that cost accelerates for a reason visible directly in the closed-form spread formula. A desk tuning the model’s risk-aversion parameter is moving along a convex cost curve with an interior efficient point, and the paper’s own data puts that point in the middle of its tested range, at neither end. Every practitioner explainer reproduces the formula. None of the ones surveyed for this piece reproduces the bill.

What the formula solves

Avellaneda and Stoikov’s 2008 paper sets up a dealer holding inventory q in a stock whose mid-price s follows a Brownian motion with volatility σ. Buy and sell orders arrive at the dealer’s quotes as a Poisson process, with intensity decaying exponentially in the distance δ from the mid-price. Solving the dealer’s expected-utility maximization gives a reservation price

where γ is the dealer’s risk aversion and T − t is the time left in the session. A long position (q > 0) pulls the reservation price below the mid, tilting both quotes down so the dealer is more likely to sell back toward flat. The optimal total spread is

where k governs how quickly fill probability decays with distance from the mid. This is the formula reproduced, with only cosmetic variation, across the current crop of practitioner explainers, including the Hummingbot market-making guide and the HFT Book reference page that calls it “the canonical model for optimal market-making quotes under inventory risk.” Neither site is used below for any figure, only for how the model is currently presented.

The founding paper’s own numbers

Avellaneda and Stoikov tested their inventory strategy against a symmetric benchmark that quotes the identical spread but centers it on the mid-price instead of the reservation price, running 1,000 simulated price paths for three levels of risk aversion (s = 100, T = 1, σ = 2, k = 1.5, A = 140 held fixed throughout). Their published tables give the following mean profit and profit dispersion across those paths.

The inventory strategy never wins on mean profit, and that is by design, since it is buying something else: the standard deviation of final inventory fell from 8.66 to 2.80 shares at γ = 0.10 and from 9.06 to 1.88 at γ = 0.50. The symmetric quoter earns more for a mechanical reason. It quotes tighter to the mid on average and fills more often, and when every counterparty is an uninformed Poisson arrival, more fills at a given spread means more expected profit.

The number worth staring at is the jump between rows. The cost of skewing runs 0.86%, then 6.35%, then 48.76%. Dividing each strategy’s mean profit by its own standard deviation, a rough profit-to-dispersion measure that is this piece’s computation and should not be read as an annualized Sharpe ratio, the inventory strategy’s advantage over the symmetric one is 1.52x at γ = 0.01, widens to 2.14x at γ = 0.10, then narrows back to 1.58x at γ = 0.50. Risk-adjusted efficiency peaks in the middle of the tested range. More risk aversion is a better trade than less up to a point, and past that point it is a worse one.

Where the convexity comes from

The acceleration is visible in the spread formula itself. The inventory-risk term γσ²(T − t) scales linearly in γ. The fill-calibration term (2/γ)ln(1 + γ/k) shrinks as γ grows: with the paper’s own k = 1.5, it runs from 1.33 at γ = 0.01 down to 1.15 at γ = 0.50. Avellaneda and Stoikov’s own Tables 1 to 3 report a single “Spread” value per γ, 1.33, 1.29, and 1.15, matching this fill-calibration term exactly; the decomposition below, including the full t = 0 spread in the last column, is this piece’s own computation from the stated formula.

As γ rises, essentially all of the additional spread comes from the inventory-risk term. Because fill intensity decays exponentially with distance from the mid, a spread that widens linearly in γ suppresses the fill rate, and therefore profit, at an accelerating rate. That is the whole mechanism behind the 0.86% to 48.76% jump. It falls directly out of the closed-form solution once its two additive terms are separated, and it means the risk-aversion parameter is a position on a convex cost curve, with an interior optimum that has to be located per instrument, since it moves with σ and k.

There is a second gap between the formula and its own foundations, on timing. Guéant, Lehalle, and Fernandez-Tapia’s 2013 paper solves the same control problem exactly, with a hard inventory limit, and finds the true optimal quotes barely move with time until the session nears its terminal point; the 2008 closed-form quotes are the special case of that exact solution valid only when T − t is small. A formula calibrated, mathematically, for the last stretch of a session is nonetheless what at least some current implementations run across the whole day: the Hummingbot guide linked above implements the plain 2008 formula with no time-independence correction. Whether institutional systems share that pattern is unverifiable from public data, and a 2022 academic study states that extensions in the 2013 lineage are already in use at major market-making firms, so the gap likely sits between open-source tooling and institutional practice rather than inside institutions.

The cost is real in live data too, and it may not be unavoidable

The same 2022 study backtested the model on 30 days of real BTC-USD Level 2 data. A genetically calibrated, static-parameter Avellaneda-Stoikov model produced by far the lowest mean and standard deviation of maximum drawdown among five tested models, while two reinforcement-learning variants that were allowed to deviate from its risk-minimizing quotes beat all three baselines, the pure model included, on Sharpe ratio on 24 of 30 days and Sortino on 25 of 30, at the price of occasional outlier drawdowns. That is the 2008 tradeoff reappearing in live data on the authors’ own reading: a correctly calibrated Avellaneda-Stoikov quote is the risk-minimizing choice by construction, so anything chasing higher risk-adjusted return is accepting more risk somewhere. It is one study, one asset pair, one month, and should be read as a documented data point rather than a settled pattern.

Whether the profit cost is a permanent law of quoting against inventory is a separate and open question. The 2008 model assumes no counterparty knows anything the dealer doesn’t, and early evidence suggests that assumption is what generates the cost: a 2025 preprint by Barzykin, Bergault, Guéant, and Lemmel (not yet peer-reviewed, with one author at HSBC under an HSBC-funded initiative) finds in its numerical illustrations that once order flow is modeled as partly informed, quotes that account for that information can raise expected profit and cut its variance simultaneously relative to informationally naive quoting. If that result survives peer review and replication on live data, the tradeoff documented above is a consequence of one dropped assumption, with a published correction. Until then, it stands as the model’s own arithmetic.

What would change this view

The 0.86%, 6.35%, and 48.76% figures are specific to the paper’s toy parameters (s = 100, σ = 2, k = 1.5, A = 140) and will not reproduce numerically elsewhere. The structural claim generalizes: the cost of risk aversion is convex because one spread component scales linearly in γ while the other shrinks, and that follows from the formula, independent of calibration. What would overturn the practical conclusion is a demonstration that, for realistic calibrations, the interior efficiency peak sits so close to one end of the usable γ range that treating the parameter as a one-directional safety dial loses nothing. The three published points here show a clear interior peak; a denser sweep on a real instrument could in principle show otherwise, and that sweep is exactly the check recommended below.

What to check

Two questions for any system running this model. First, where does the current γ sit on its own cost curve: sweep γ against the desk’s actual σ and k calibration, compute mean profit and its dispersion per setting the way the 2008 paper did, and locate the interior peak instead of assuming more risk aversion is monotonically safer. The founding paper’s own three data points show it is not. Second, is the system running the 2013 exact quotes or the 2008 asymptotic ones, since the latter are accurate only near the session close and the difference is a documented, closed-form correction, published for over a decade. Both checks run in hours against a calibration the desk already owns, and each prices something currently being paid for without being measured.

📊 Want Deeper Quantitative Analysis?

This research took a long stretch of data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon. By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

You can also follow my work here:

• YouTube: The Mathematical Trader

• LinkedIn: Navnoor Bawa

Europe’s Markets in Crypto-Assets Regulation excluded the world’s largest stablecoin from every licensed venue in the bloc, the textbook setup for a regulatory-segmentation premium of the kind Korean traders have traded for a decade. That premium has not shown up in spot prices, at least not in the most recent data this piece could verify. The reason is more useful than the premium itself would have been. The capturable edge here is a monopoly rent on stablecoin infrastructure, captured through Circle’s equity. Two things bound it: Circle’s own trading liquidity, which caps a same-day position near 150 million dollars, and two independent decay clocks that a position sized against this thesis needs to respect separately.

🎬 Prefer to watch rather than read? A NotebookLM-generated video overview of this article is available here: Watch the video overview → Full analysis, citations, and data remain in the article below.

The pattern this looks like

A rule that legally locks an asset out of a regional market while global demand for that asset stays intact is the standard precondition for a law-of-one-price violation. South Korea has run this experiment for a decade. Bloomberg reported that capital controls drove the “kimchi premium,” the gap between bitcoin priced on Korean exchanges and everywhere else, to 51 percent in January 2018 before it collapsed over seven weeks. The academic literature on the mechanism is worth being precise about, because the precision itself is a lesson. The founding paper on the subject, Choi, Lehar and Stauffer on SSRN, reports in its current working-paper draft that bitcoin traded on average 2.27 percent higher in Korea than the US between January 2016 and January 2020, a lower figure than the 4.73 percent average and 54.48 percent January 2018 peak that circulate widely in secondary coverage. Those secondary figures trace to an earlier 2018 draft of the same paper, using a shorter sample window; the extended sample in the current draft pulls the average down because the post-2018 years ran cooler. The paper itself is still listed by its own author as revise-and-resubmit at the Journal of Financial and Quantitative Analysis. It has not been accepted or published, so “peer-reviewed” would overstate its status; both sets of figures are working-paper estimates that have not been through journal refereeing. A second, separate paper, Seo, Koo and Yang in Economic Modelling, models the premium with threshold regression and finds it behaves as a near-random walk at small magnitudes, where arbitrage does not clear it, and mean-reverts only once it grows large enough to draw capital despite the friction. The paper also estimates a non-zero long-run steady-state premium of 1.24 percent for bitcoin: the law of one price never fully holds even at rest. Won convertibility restrictions are the binding constraint behind all of this. A foreign trader cannot simply wire funds in to close the gap, which is why the premium has resurfaced repeatedly since 2018 despite years of tightening compliance, including a 20.8 percent spike in May 2021 and a 3-to-5 percent spike amid political turmoil in December 2024.

MiCA’s e-money token rules impose an analogous asymmetry: a specific asset, USDT, is legally barred from every MiCA-licensed venue while nothing stops non-EU demand for it. If segmentation alone produced a premium, this is where a professional would expect to find one.

Why the price-arbitrage version of this looks dead, with a caveat on how current the evidence is

The clearest data this piece could verify comes from Kaiko’s “State of the European Crypto Market” report, cited in a CryptoSlate analysis: as of November 2024, 30-day average bid-ask spreads on the tightest EU venues sat at 2.6 basis points on Bitvavo and 3 basis points on Kraken, in the same range as global majors. That is the opposite of what Korea looked like at any point its premium exceeded single digits. The caveat has to be stated plainly: November 2024 predates the July 2026 deadline by about a year and a half, and this piece could not locate a more recent, directly comparable spread dataset. The claim this evidence actually supports is narrower than it might look: EU spot markets were competitively priced heading into the deadline. Whether they remain so today carries considerably less certainty.

A separate, single-sourced data point points the same direction. The Block reported a Kaiko statement that MiCA-licensed exchanges carried roughly 83 percent of European trading volume as of June 2026, ahead of Binance’s unlicensed exit. This piece could not locate the underlying Kaiko publication behind that specific figure, only The Block’s citation of it, so it is treated here as a media-relayed statistic attributed to a credible named source, without independent verification. Taken at face value, it explains the mechanism if the spread data is also taken at face value: the flow that would have been captive and price-insensitive, the way Korean retail was captive to won-denominated exchanges, was already routed through licensed venues before USDT lost its listing there. No large pool of forced buyers remained to bid prices away from the global level. Treating a spot-price kimchi-style trade on EU crypto as a live, current edge, on evidence that is partly a year and a half old and partly a single unverified citation, would overstate this piece’s confidence. The more defensible claim is narrower: there is no verified evidence of one, and the burden sits with anyone asserting the trade exists to produce more current spread data.

The mechanism that is real: a rule that forces a global decision

Korea’s premium is a currency-convertibility problem. MiCA’s is a reserve-composition problem, and the two propagate differently. Under MiCA’s e-money token regime, an issuer must hold at least 30 percent of reserves as deposits at EU credit institutions, rising to 60 percent for tokens the European Banking Authority designates significant, confirmed by an academic account of the regulation published by Springer and by legal analysis from Ramparts. Significance is triggered under Article 43 by criteria including more than 10 million holders, a market capitalization above 5 billion euros, or average daily activity above 2.5 million transactions and 500 million euros in value, both conditions applying together. USDT, with a market capitalization near 184 billion dollars, a number-three ranking among all crypto assets, and several hundred million users, clears every one of these thresholds by orders of magnitude.

The reserve requirement attaches to the token itself, wherever it circulates. Because USDT is a single fungible instrument that settles peer to peer on public chains with no way to tag which units sit with EU holders, there is no way for Tether to apply the 60 percent EU-bank-deposit rule to only its EU-facing float. Compliance would mean restructuring the reserve backing the entire outstanding supply, moving well over 100 billion dollars of it into deposits at European banks whose deposit insurance caps out at 100,000 euros per institution. CEO Paolo Ardoino has called that structure very dangerous when it comes to stablecoins, and in a separate interview pointed to Circle’s own exposure during the 2023 Silicon Valley Bank failure, saying Circle “almost died” in that episode, as the scaled-up version of the risk he is refusing to take on. That episode is well documented: Circle’s own account and a Federal Reserve Board research note both confirm that 3.3 billion dollars, about 8 percent of USDC’s reserves, was briefly stranded at the failed lender before being made whole days later. Circle made the opposite calculation from Tether, securing an Electronic Money Institution license from France’s ACPR that authorizes both USDC and its euro token EURC across all 27 member states.

The all-or-nothing structure is confirmed by what smaller issuers have done instead of converting a global token: StablR issues EURR and USDR, new EU-specific stablecoins built on Tether’s own Hadron tokenization platform, with payments network Oobit integrating them for merchant use instead of issuing them itself, ring-fenced from the start. Tether will apparently help other issuers build a segregated compliant product; it has not done this for USDT itself. That is the detail that separates an economic constraint from a branding decision, and it is the detail the decay analysis below turns on.

What the rent is actually worth, and what this piece cannot claim about it

Circle’s own numbers show what a rent of this kind looks like at scale, though this piece cannot isolate the slice specific to USDT’s EU exclusion. Reserve income, the interest Circle earns on the assets backing USDC and EURC, ran to 733 million dollars in the fourth quarter of 2025 alone on 75.3 billion dollars of USDC in circulation at year end, per Circle’s own results; full-year 2025 revenue and reserve income totaled 2.7 billion dollars. Circle does not break this out by jurisdiction, and the growth reflects several drivers beyond MiCA, including the US GENIUS Act, new blockchain integrations, and the Circle Payments Network, so no figure here isolates a specific EU rent this piece can point to. What the numbers do establish is the order of magnitude the mechanism operates at: reserve income scales directly with circulating USDC and EURC, and every dollar of stablecoin balance that would otherwise sit in USDT and now sits in a MiCA-compliant token adds directly to that base. The word “rent” in this piece describes that structural relationship; this piece has not measured it in dollar terms.

Decay and capacity

Two decay tracks run independently, and a position built on this thesis needs to be sized against the faster one.

The regulatory track has a fixed date. The European Commission opened a public and targeted consultation on May 20 to assess whether MiCA remains fit for purpose, with feedback open until August 31 and the treatment of multi-jurisdictional stablecoin issuance explicitly in scope. OMFIF’s Orchard told CoinDesk that the Commission is reportedly weighing a shift toward a GENIUS-Act-style reserve model, letting issuers hold European government money-market instruments instead of bank deposits. That characterization is an analyst’s read on a pre-decisional process. The Commission has not confirmed it as its position, and it should carry the caution any pre-decisional policy read deserves.

Set against that, the fungible-token mechanism argues the exclusion is more likely to deepen than dissolve even if the reserve percentage is eased. Tether’s outstanding supply sat near 184 billion dollars as of early July 2026; every additional billion in global float raises the absolute size, and therefore the concentration risk, of whatever share MiCA would require in EU bank deposits. A rule Tether could plausibly have absorbed at a 10 or 20 billion dollar scale becomes proportionally harder to accept as the balance sheet it would apply to compounds. Regulatory easing on the margin, a lower percentage, or government paper instead of bank deposits, softens the mechanism; it does not reverse the growth dynamic working against reversal. That is the non-obvious part of this thesis: the exclusion is a rule interacting with a balance sheet that keeps getting larger, and that interaction pushes reversal further away over time, independent of what the August consultation concludes.

The competitive track runs on its own clock and does not depend on Brussels at all, and here the mechanism is more tangled than a clean regulatory story. Circle’s shares fell by a reported 15 to 18 percent in a single session in the days around June 30, depending on the source and the exact measurement window (Yahoo Finance put it near 16 percent intraday; crypto.news and a Bernstein note cited by Yahoo Finance put it at 17.5 percent). Reporting attributes that move to two compounding causes: Circle’s removal from several Russell indexes on June 26 forced mechanical selling from index-tracking funds, and a bank-and-payments consortium’s launch of a rival stablecoin, Open USD, on June 30 raised competitive doubts about Circle’s reserve-interest revenue model at the same time. Disentangling how much of the drop was passive rebalancing and how much was a fundamental repricing is not possible from public reporting; this piece did not find a source that isolates the two. Shares partially recovered in the following days, trading back into the mid-60s per Yahoo Finance, though this piece could not verify a precise single-day rebound percentage from a source stronger than a promotional crypto-market blog, so none is given here. The exact percentage matters less than the underlying fact: OUSD is a global product. It attacks Circle’s reserve-income model from outside the regulatory perimeter Circle currently has largely to itself in Europe, so Circle’s position can be eroded by competition even if the MiCA-specific regulatory moat holds exactly as it stands today.

Capacity for expressing this thesis through Circle’s equity is bounded by the stock’s own trading liquidity, a far tighter constraint than the size of the underlying stablecoin market it is contesting. Circle traded at 65.92 dollars on July 2 against a 15.19 million share average daily volume, per Robinhood’s market data; this piece relies on a single retail-brokerage data source for that specific figure and treats it as an indicative figure only. Holding a single day’s participation to 15 percent of that average, a standard ceiling for keeping market impact from compounding, caps same-day accumulation at roughly 2.28 million shares, or about 150 million dollars notional (2.28 million shares multiplied by 65.92 dollars). Repeating that same 15 percent daily ceiling across a five-trading-day week, and assuming no other buyer is competing for the same size, scales the accumulated position to roughly 750 million dollars (five days multiplied by 150 million dollars). Above that range, a position sized against this specific thesis starts to become a meaningful fraction of the float itself, at which point the trade’s own execution risk competes with the regulatory and competitive risks it is meant to isolate.

The objection: could Tether just ring-fence a compliant token overnight

The Hadron-based precedent set by StablR shows this is technically available to Tether on a timescale of months, which is the strongest argument against treating the exclusion as durable. That precedent needs a caveat a professional would want before leaning on it: on May 24, 2026, StablR’s EURR and USDR both depegged sharply, EURR falling roughly 26 percent and USDR roughly 36 percent, after an attacker compromised a single key on a 1-of-3 minting multisig and printed about 13.5 million dollars in unbacked tokens; StablR acknowledged the exploit in a public statement roughly eight hours after the on-chain activity had stopped. The failure traces to governance and key management. It leaves the token-design mechanism this piece describes intact, so it does not undercut the technical-feasibility point, but it does mean the flagship example of fast, low-friction ring-fencing had a real security failure weeks before this analysis was written. That should discount how much confidence a reader places in a fast compliance path as a clean, low-risk one. The answer to the objection itself is that Tether has had the option to do this since the StablR and Oobit partnerships launched and has not used it for USDT itself, choosing instead to let third parties issue separately branded compliant tokens. A ring-fenced “USDT-EU” would fragment USDT’s own liquidity and network effects, the asset’s core commercial value, in exchange for solving a problem Tether’s leadership has publicly framed as Europe’s to fix. That is a strategic choice with its own switching cost, and Ardoino’s public position gives no indication that calculation is close to flipping. Reversal remains possible in principle. But the path runs through a corporate decision Tether has had the tools to make for over a year and has not, which is a weaker basis for expecting near-term change than a pure cost-of-compliance argument would suggest.

What would change this view

Three observations would falsify the “no live price arb” half of this thesis, and a professional acting on this piece should look for current data on all three before relying on it: a sustained widening of EU-venue bid-ask spreads relative to global majors beyond the 2.6-to-3 basis point range Kaiko measured in November 2024, a drop in the MiCA-licensed share of EU volume materially below the 83 percent single-sourced figure cited above, or persistent EUR-denominated pricing on licensed venues running above transaction costs relative to global benchmarks and above the 1.24 percent steady-state background level the threshold-regression paper finds even at rest, the signature the kimchi-premium literature associates with a genuine, arbitrage-resistant gap. This piece did not find current data confirming or ruling out any of the three with the precision the claim deserves. That gap should be read as a limitation of the available evidence.

The regulatory-rent half of the thesis would be falsified two ways. The European Commission’s review could conclude with the significant-EMT threshold left untouched or tightened, which would confirm the exclusion as durable. Or Tether could announce a ring-fenced MiCA-compliant version of USDT itself, which would undercut the corporate-inertia argument directly.

The actionable read, and its size

There is no verified, current EU-specific spot-price arbitrage to chase in bitcoin or ether. The best available evidence, some of it dated, says that gap was closed or never opened before MiCA’s deadline arrived, but the evidence is not current enough to state that with full confidence. The better-supported tradeable idea is a bet on Circle’s regulatory position, understood as a structural advantage this piece has not sized in dollar terms. Position size should track Circle’s own liquidity, since the stablecoin market it is contesting is far larger than the equity float that trades on it: same-day entries capped near 150 million dollars, a full position capped near 750 million dollars if built over a week, reviewed against two independent triggers, the European Commission’s consultation closing August 31, and any signal that Tether is willing to ring-fence USDT itself instead of leaving that work to third parties. A position held past either trigger without being re-underwritten, or one sized on the assumption that the spot-arbitrage evidence in this piece is more current than it is, is a bet on inertia and stale data. The mechanism this piece describes does not support holding past that point.

A fuller institutional version of this thesis, with four dated catalysts, bull, base, and bear price scenarios with the underlying math shown, named position sizing against the same liquidity constraint, and a complete verification log checking every figure here against a primary source, is available at the full note on Patreon.

Deeper research like this lives on Patreon

This piece went through four rounds of correction and verification against primary sources, including cross-checks against two independent fact-check reports. The same standard applies to the trade notes, mechanism explainers, and institutional-grade research published there regularly. Joining supports that work directly and funds more of it.

→ Join the Patreon community

Also on YouTube and LinkedIn.

On February 5, 2018, the S&P 500 fell 4.2%, a move the Bank for International Settlements measured at 3.8 standard deviations. Under the normal distribution most parametric risk models implicitly assume, a move that size happens about once every 55 years; under Chebyshev’s inequality, the distribution-free descendant of Markov’s inequality that assumes nothing about the shape of the return distribution at all, the same event carries a ceiling near 6.5%, roughly 895 times looser. That gap is not a historical curiosity. The specific products that failed that day (XIV, and the -1x version of SVXY) are gone or delevered, verified by the issuer’s own SEC filing within three weeks of the event. But the underlying condition that produced the gap, parametric models calibrated during a period of unusually low realized volatility, sitting under leveraged, mechanically-rebalanced, negative-convexity exposure, now shows up in a market that scarcely existed at the time and has since become one of the largest corners of US options trading.

The consensus is correct, as far as it goes

Ask any desk why nobody uses Markov’s inequality for risk management and the answer is fair: it is famously, almost comically loose. A 2023 paper working through the traditional Markov bound against the true tail of an exponential distribution found it overstating the tail probability by roughly 145 times at the 99.9th percentile, and by several orders of magnitude further into the tail. Stanford’s own introductory treatment of the inequality makes the same point with a coin-flip example: Markov’s bound puts the probability of 16 or more heads in 20 flips of a p = 0.2 coin at 25%, when the true probability is close to 1.4 x 10⁻⁸, a bound about 18 million times too wide. That is the price of a result that has to stay valid for every distribution consistent with a given mean, however strange, however fat-tailed. Nobody sizes a book off a number that far from the truth, and nobody should. The critique is correct.

Why the loose bound is the point

Markov’s inequality requires the variable to be non-negative: for X ≥ 0 and a > 0, P(X ≥ a) ≤ E[X]/a. Returns are not non-negative, which is the first objection a sophisticated reader should raise. The fix is not to wave the requirement away, it is to apply the inequality to the loss, L = max(0, -R), a non-negative quantity by construction. Markov’s inequality then bounds P(L ≥ a) directly, using nothing but the expected loss: no variance, no distributional shape, no model.

Add variance and the bound sharpens considerably. Squaring the deviation from the mean and applying Markov to (X - μ)² produces Chebyshev’s inequality, P(|X - μ| ≥ kσ) ≤ 1/k², a derivation both Stanford’s notes and a University of Washington concentration-inequalities course walk through identically. A one-sided refinement credited to Cantelli in 1928 sharpens this further for a purely one-directional loss event: P(X - μ ≤ -kσ) ≤ 1/(1 + k²).

What matters here is not the algebra, it is the decay rate. A Gaussian tail collapses exponentially in k², so by k = 4 or 5 a parametric model assigns a probability that is effectively zero. Chebyshev and Cantelli decay only polynomially, in 1/k², because they must stay valid for distributions that do not thin out that fast. At k = 2 the two views sit roughly an order of magnitude apart. By k = 4 they are separated by more than three orders of magnitude. The gap does not grow with the size of the move, it grows explosively with it, which means the two frameworks disagree most sharply in exactly the regime where a parametric model’s tail assumption is doing the most unverified work.

February 5, 2018, in the numbers

Short volatility had been one of the decade’s quietest trades. The VIX closed at an all-time low of 9.14 on November 3, 2017, and VelocityShares’ inverse VIX note, XIV, rose from $6.51 at the end of 2011 to $134.44 at the end of 2017, a 20x return, per a market commentary hosted by Cboe on its research-publications page; the document itself is authored by a third-party asset manager, DGV Solutions, and is self-labeled as commentary rather than research, so it is cited here only for objective price history, not for its own market views. Assets across the small group of leveraged and inverse VIX exchange-traded products reached roughly $3.5 billion by early February 2018, per a peer-reviewed study in the Financial Analysts Journal by Augustin, Cheng, and Van den Bergen, a figure consistent with BIS’s own independent estimate of roughly $4 billion at end-2017 once smaller products beyond XIV and SVXY are included.

On the day itself, the S&P 500’s 4.2% decline was the 3.8-standard-deviation move BIS later calculated, and the VIX moved from 17.31 to 37.32, a 20-point, 115.6% jump, the largest single-day percentage increase in the index’s history and more than double the prior record of 64.2% set in February 2007, again per the Cboe-hosted DGV commentary cited above; BIS corroborates the order of magnitude independently, describing the same move in its own prose as roughly a 20-point jump and the largest daily VIX increase since the 1987 stock market crash, without restating the precise decimal figures. A peer-reviewed analysis in the MDPI journal Risks put a number on the surprise: given the VIX’s typical -0.8 historical correlation with the S&P 500, a 4% equity decline should have produced a roughly 3.2-point rise in the VIX, not a 20-point one.

The mechanism was structural, not purely sentiment-driven. BIS’s own transaction-level analysis documents that both long and short volatility ETPs needed to buy VIX futures near the 4:15pm close to maintain their target exposure, a rebalancing collision that pushed 115,862 futures contracts, roughly a quarter of the day’s entire volume, through the market within a single minute at 16:08. The Augustin, Cheng, and Van den Bergen study adds the input that made this collision so dangerous in the first place: between 2007 and 2017, the S&P 500 VIX Short-Term Futures Index carried an average 90-day trailing volatility of 64.0% (the spot VIX itself averaged 117.2%), versus 17.4% for the S&P 500, and by late 2017 the S&P 500’s own trailing volatility had compressed further, to about 6.8%. Any model borrowing equity-scaled tail assumptions for this product class, calibrated during an unusually calm stretch, was miscalibrated before the shape of the tail even entered the picture.

The value of XIV fell 84% during the regular session and the product was terminated, triggering the acceleration clause in Credit Suisse’s own prospectus, which permitted termination once the note’s intraday indicative value fell to 20% or less of the prior day’s close. The closing indicative value on February 2 had been $108.3681. That 84% figure is BIS’s own number for the regular session specifically; by the time VIX futures finished spiking past 50 in after-hours trading, Six Figure Investing puts the full-day collapse at 97% for XIV, 91% for SVXY, and 87% for VMIN, the fuller unwind that the acceleration clause was ultimately settled against.

The other side of that trade has a name. Set Capital LLC v. Credit Suisse Group AG, a securities class action, alleges that Credit Suisse’s own hedging of its XIV exposure helped manufacture the liquidity squeeze that crashed the notes, clearing the way for the bank to redeem them at the crashed price. The Second Circuit revived the market-manipulation claims in 2021 after a district court had dismissed them, finding the allegations plausible enough to proceed; the investors’ own counsel puts Credit Suisse’s resulting profit at $475 million to $542 million, and a federal judge granted partial class certification on the manipulation claims in February 2025. The case has not gone to trial and the allegations are unproven. What is not in dispute is the shape of the trade: someone was structured to gain close to what retail noteholders were structured to lose, and it was the bank that sold them the note.

The gap, computed

Set the litigation aside and return to the number that started this. Take BIS’s 3.8 figure and run it through both frameworks. Under a normal distribution, the one-sided probability of a move that size or worse is about 0.0072%, or roughly one in 13,800 trading days. Under the Cantelli bound, which requires only the mean and variance and assumes nothing about shape, the ceiling on the same event is 6.48%, or roughly one in 15 observations. The Cantelli ceiling is about 895 times the Gaussian point estimate.

Run the same exercise one level down, on the VIX futures index itself, using Augustin, Cheng, and Van den Bergen’s own 64.0% realized-volatility figure. Converting that to a daily figure (dividing by the square root of 252 trading days, the standard annualization convention) gives a daily volatility near 4.0%. XIV’s 84% one-day fall implies, under a -1x daily-rebalanced product where a single day’s fund return approximates the negative of the index’s return, that the underlying futures index itself moved by a comparable magnitude that day, roughly 21 standard deviations by this measure. At that point the Gaussian framework does not just underestimate the tail, it stops producing a usable number: the implied one-sided probability is on the order of 10⁻⁹⁴ percent, a figure with no operational meaning. The Cantelli ceiling, built from the same inputs, still returns something a risk committee could act on: about 0.23%, or roughly one in 435. This is the sharper version of the same point. A distribution-free bound degrades gracefully as an event gets more extreme. A parametric one does not degrade, it fails silently, producing a confidently-stated number that has stopped meaning anything long before a risk manager would notice.

This is worth running as a live check, not only in hindsight. Cboe’s own VIX data shows the index closing at 16.45 on July 1, 2026, the day before this piece was finalized. Using implied rather than realized volatility as the input this time, match Volmageddon’s exact magnitude, a 4.2% one-day S&P 500 decline, against that single number, and the move works out to roughly 4.05 standard deviations: slightly more extreme than the 3.8 BIS calculated from realized volatility for the actual 2018 event. The Cantelli ceiling on a move that size is about 5.7%; the Gaussian estimate is about 0.0025%, a gap of roughly 2,270 times. The exact multiple will be different by the time this is read, since implied volatility moves daily. That is the point: the check costs nothing more exotic than the VIX print and can be rerun every morning, which makes it a live number rather than a 2018 artifact.

Neither of these ceilings is the true probability. What the gap measures, at any of these levels, is how much of a model’s stated confidence comes from something verified in the data (the mean and variance) versus something assumed on top of it (the tail’s shape). The only thing standing between “basically impossible” and “small but real” is a curvature assumption nobody in the chain re-verified before February 2018.

Why this is a decayed vehicle, not a decayed mechanism

An honest reading of this case has to ask whether the industry already fixed it. It largely did, for the specific products involved. ProShares’ own SEC filing, dated February 26, 2018, three weeks after the event, announced that SVXY’s target exposure would be cut from -1x to -0.5x and UVXY’s from 2x to 1.5x, effective the next trading day. XIV was terminated outright. VMIN, a smaller competitor, was wound down by November 2018 for lack of assets. Vanguard, Fidelity, and Merrill Edge each restricted retail access to leveraged and inverse volatility products within the following year. A simulation from Six Figure Investing, cited above, estimated that a repeat of the February 5 move against the post-delevered SVXY would produce roughly a 48% loss today, severe but survivable next to the 87% to 97% full-day losses realized in 2018. Current fund-data snapshots vary by provider and update lag: a same-day check of TradingView puts SVXY at $218.21 million, while Yahoo Finance and U.S. News, sourced to Morningstar, have shown figures ranging from roughly $190 million to $250 million across recent weeks, and Danelfin puts SVXY’s -1x successor SVIX in the neighborhood of $190 million to $211 million depending on the snapshot date. Combined, that is somewhere in the $400 to $460 million range, against a 2018 peak of roughly $3.5 to $4 billion across XIV, SVXY, and VMIN depending on the cutoff date and which products a given source includes, a reduction on the order of 87 to 90% under any reasonable reading of the range. Judged purely on the 2018 vehicle, this is a decayed setup: smaller, less leveraged, harder for a retail account to reach, and already the subject of a peer-reviewed forensic account with “crowded trades” in its own keyword list, on top of active federal litigation over who profited from it.

The mechanism is a different question, and whether it can recur in a new vehicle is not settled, it is actively and specifically contested. Distribution-free tail checking is not a crowded trade in the conventional sense, because it is not a return-generating signal that gets arbitraged away as more capital adopts it; running it does not consume anyone else’s ability to run it. So the relevant capacity question is not the AUM at which this stops working, it is how much capital currently sits in structurally similar exposure priced by models with the same blind spot. Zero-days-to-expiration options share the 2018 products’ core ingredients (leverage, mechanical hedging by market makers, and pricing built on realized volatility from unusually calm recent windows), and they have grown from 21.5% of total US listed options volume in 2024 to 24.1% in 2025, citing OCC and Cboe data, reaching more than half of all S&P 500 index options volume by the fourth quarter of 2024. Whether dealer hedging in that market can produce a Volmageddon-style collision, not merely share its ingredients, was tested in public on August 15, 2023, when the S&P 500’s decline accelerated by roughly 0.4% in twenty minutes. Goldman Sachs’s Scott Rubner attributed the acceleration to 0DTE-driven dealer hedging, as Bloomberg reported the next day. Cboe’s own gamma-exposure reconstruction of that same afternoon found market makers net long gamma, meaning their hedging should have dampened rather than amplified the move, until 3:30pm, by which point the index had already stabilized, and concluded the data does not support 0DTE hedging as that day’s driver. Two data-facing institutions read the same twenty minutes and reached opposite conclusions. Bank of America’s Nitin Saksena has separately argued the risk is overstated, while relaying an alarm raised by others that positioning could produce an event echoing Volmageddon; in the same reporting, J.P. Morgan’s Marko Kolanovic warned that 0DTE-driven swings could reach $30 billion in a single session. None of this proves the mechanism will recur at Volmageddon scale in this market. It shows the mechanism is contested by name, in public, by desks with access to the actual positioning data, inside a market that is now a majority of all S&P 500 options volume, on a scale the 2018 VIX-ETP complex, a few billion dollars in assets at its peak, never approached.

This isn’t a new idea. It’s an abandoned one.

Distribution-free, worst-case tail reasoning is not a novel proposal for portfolio construction, it is close to where modern portfolio theory began. A.D. Roy’s 1952 paper, “Safety First and the Holding of Assets,” proposed choosing a portfolio to minimize the probability that its return falls below a disaster level, d. Because Roy did not want to assume a specific return distribution, he bounded that probability using a one-sided form of Chebyshev’s inequality, which reduces the minimization problem to maximizing (μ - d)/σ, the same algebraic form, over a decade before it would be popularized as the Sharpe ratio.

Roy’s paper appeared the same year as Markowitz’s “Portfolio Selection.” Harry Markowitz himself, in a 1999 historical essay quoted in Mark Rubinstein’s 2002 retrospective in the Journal of Finance, described his own “father of modern portfolio theory” title as one Roy deserved to share equally. What separates the two founding contributions is not the ratio, which converges to the same form, it is the justification underneath it. Markowitz’s framework optimizes over an assumed joint return distribution. Roy arrived at an identical-looking ratio specifically to avoid needing one. Over the decades since, the industry kept the form (every risk-adjusted-return metric from Sharpe to Sortino to the information ratio is a variation on (numerator - benchmark)/dispersion) and let the distribution-free caution that originally justified interpreting a given number of standard deviations fall away, replaced by parametric or historically-simulated confidence intervals that require exactly the assumption Roy built his ratio to avoid needing.

Three objections, all worth taking seriously

The first: a 6.5% ceiling is too loose to act on, and nobody can post capital against a bound that wide without crippling the book. This is correct, and it is not the proposal here. Using the Cantelli ceiling as a replacement point estimate for VaR would produce a worse model, not a better one, for the same reason the consensus section above concedes. The actionable version is narrower: track the ratio between the parametric estimate and the distribution-free ceiling for any book carrying negative convexity, not the level of either number alone. Del Castillo’s 2023 paper makes the relevant point about why the ratio, not the raw bound, is the useful object, noting that avoidance of model risk is decisive when multiple competing models are present in a real-world situation. A ratio that stays in the single digits says the parametric model’s tail assumption is doing modest work. A ratio in the hundreds says the position’s entire safety margin rests on an unverified assumption about curvature.

The second: safety-first portfolio construction was tried and superseded by mean-variance and expected-utility approaches for good reason, because a criterion built on a worst-case bound is systematically overcautious and leaves return on the table in ordinary markets. That is also correct, and it is not an argument for reviving Roy’s allocation framework wholesale. Mean-variance optimization dominates safety-first sizing in normal regimes, and nothing here disputes that. The claim is narrower still: not that the criterion should return, but that the specific discipline behind it, checking a claimed tail probability against what the data alone can support, was discarded along with the criterion, and nothing in modern risk architecture replaced it.

The third, and the one a reader of the decay section above should raise directly: this whole case is 2018 news, already fixed by the very deleveraging and product terminations documented three sections up. Partially, and that partial concession is the point rather than a weakness in it. The specific instrument decayed. The condition that produced the gap, a parametric tail assumption unverified against the position’s own dispersion, is a property of how risk gets modeled, not of any single ticker, and the 0DTE evidence above, contested as it is, shows that condition currently sits under a larger, faster-growing, and actively disputed book of exposure than the one that broke in 2018. A technique with no crowding mechanism does not become less useful because the specific product it was first demonstrated on stopped trading.

What would change this view

Two things would weaken this argument considerably. First, public filings cannot show whether a given fund already runs a check like this. Quarterly 13F filings disclose long, US-listed equity positions only, as of quarter-end, on a 45-day reporting lag; they exclude derivatives, short positions, and non-US holdings entirely, which means the leveraged and short-vol exposure this piece is about would never appear in one regardless of which fund holds it. No standard SEC disclosure type, 13F included, reveals internal risk methodology. If distribution-free floors are already standard practice inside systematic multi-strategy risk committees, or inside the market-making desks now absorbing the bulk of 0DTE flow, the gap described here is being managed quietly rather than sitting unused, and the piece’s urgency, though not its math, would need revising.

Second, the 21-standard-deviation figure for the VIX futures index rests on a single realized-volatility input, an 11-year average from one peer-reviewed source, applied to XIV’s return as a proxy for the underlying index’s move. A different vol window, or a direct computation from the futures index’s own tick data rather than a leveraged product’s daily NAV, could move that specific number materially, though it would need to move by several full orders of magnitude to change the qualitative conclusion that a Gaussian framework produces an uninformative answer at this level. The equity-level 895x figure rests on firmer ground, a single BIS-calculated input with no proxy step, and is the number this piece’s actionable claim should be weighted toward. And the Goldman-versus-Cboe dispute over August 2023 cuts both ways by design: it is evidence the mechanism is live enough to argue about, not evidence it will replay at 2018 scale.

The actionable version, and its capacity bound

For any book carrying negative convexity or embedded short optionality, compute the Cantelli ceiling from nothing but the position’s own realized variance and post it next to whatever parametric VaR or expected-shortfall figure the desk already produces. Basel’s own post-crisis review moved capital requirements from VaR toward Expected Shortfall precisely because pre-crisis models systematically underpriced tail risk, and Expected Shortfall remains a parametric or historically-simulated measure, not a distribution-free one. Watch the ratio between the two numbers, not either level in isolation. This practice has no AUM ceiling of its own, because it is a governance check, not a factor that gets crowded, but it is only worth running on exposure large enough that a Volmageddon-scale ratio would matter to the book’s survival, which in practice means any negative-convexity sleeve above the low seven figures at a systematic shop, and any book at all inside a market-making or dealer operation now absorbing 0DTE flow at more than half of S&P 500 options volume. Run that check before the tail event, the way the data already allows, not after it.

📊 Want Deeper Quantitative Analysis?

This research took a long time: data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

For more research like this, subscribe to The Mathematical Trader on YouTube, or connect with me on LinkedIn.

The practitioner Black-Scholes delta — the hedge ratio computed by substituting implied volatility into the BS formula and taking the partial derivative — systematically over-hedges S&P 500 call options and under-hedges put options relative to the position that actually minimizes P&L variance. The error follows from one structural feature of equity markets documented continuously since the 1970s: volatility and price move in opposite directions. The correction is a single-equation adjustment derivable from outputs every risk system already produces. Hull and White (2017), working with 1.3 million daily S&P 500 option observations across a data set spanning January 2004 to August 2015, find that switching to the minimum variance delta reduces out-of-sample hedging error variance by 25.7% for calls and 22.5% for puts. For deep out-of-the-money calls, the reduction reaches 42.1%. For actively traded strikes, the corrected hedge outperforms SABR stochastic volatility calibrated daily across every option maturity, using roughly one-third of one percent of the parameters. The practitioner delta remains the default output of standard risk systems. Most desks have not made the switch.

What the Consensus Gets Right

The practitioner Black-Scholes model, which prices each option at its own market-implied volatility and computes Greeks by differentiation, is not naive. It is internally calibrated to market prices by construction. A practitioner computing delta for an SPX put uses that put’s actual implied volatility, not a flat surface, so the delta reflects the option’s exact location on the skew. The implied BSM delta is better than constant-volatility BS delta for this reason.

The standard defense is clear: if practitioners already substitute market implied volatility into the BS formula, have they not already accounted for the smile? This is the question the empirical literature resolves. The answer is no, and the reason is structural rather than parametric.

The Missing Term

The practitioner BS delta is a partial derivative: it measures how an option’s price changes when spot moves while implied volatility is held fixed. The actual market value change of an option when spot moves by dS includes a second term:

dC = (∂C/∂S) · dS + (∂C/∂σ) · dσ

The practitioner delta captures only the first term. The second — vega times the concurrent change in implied volatility — is not zero for equity indices. The negative correlation between equity prices and their implied volatility has been empirically documented across every major equity index for decades, first established by Black (1976) and Christie (1982), confirmed in implied volatility terms by subsequent work. When S rises, σ falls; when S falls, σ rises.

The minimum variance (MV) delta — the hedge ratio that minimizes the daily variance of the hedged position — incorporates both terms:

Δ_MV = Δ_BS + ν × E(∂σ_imp/∂S)

where ν is the practitioner BS vega and E(∂σ_imp/∂S) is the expected change in implied volatility per unit change in spot. For equity indices this expectation is negative, making Δ_MV < Δ_BS for calls. Equity call options are systematically over-hedged when using the practitioner convention; equity put options are systematically under-hedged.

This identity is exact in a two-factor diffusion framework and an approximation under jump-diffusion or non-Markov processes. The approximation is tight for equity index options because the dominant source of variation in implied volatility is the price move, not idiosyncratic vol noise. Hull and White’s regression of implied vol changes on price changes, run across 2007 to 2015, finds that roughly 60% of the total variation in implied volatility changes for deep OTM S&P 500 calls is explained by concurrent index level changes.

The term ν × E(∂σ_imp/∂S) is the P&L contribution the practitioner delta convention ignores on every hedge rebalance. One clarification is worth stating explicitly: the 60% R² comes from a contemporaneous regression — it measures co-movement on the same day, not forecasting power. The correction does not require predicting vol changes in advance; it requires only that the historical relationship between price moves and vol moves, estimated from past data, is stable enough to serve as the expected conditional response going forward. Hull and White’s rolling estimation methodology is designed precisely to test whether that stability holds out-of-sample.

The Empirical Foundation

Precisely how the vol surface moves with spot determines the magnitude of the correction. Derman (1999), in a Goldman Sachs Quantitative Strategies paper also published in Risk Magazine, introduced the conceptual framework for equity index options: the surface can operate in a sticky-strike regime, where each fixed-strike option’s implied vol is independent of spot, or a sticky-delta regime, where moneyness determines implied vol and a spot move carries the entire surface. Neither extreme holds perfectly, but the data strongly favors models in which the surface moves with spot.

Daglish, Hull, and Suo tested multiple conventions against 47 months of S&P 500 over-the-counter consensus implied volatility surfaces from June 1998 to April 2002. The relative sticky-delta model fit the surface with an R² of 94.93% and an out-of-sample RMSE of 0.73 percentage points of implied volatility. The sticky-strike model achieved R² of 27% and RMSE of 5.25 percentage points — a 7.2× difference in out-of-sample RMSE. The paper’s F-statistic for equal explanatory power between the two models is 32.71, a decisive rejection at any conventional significance level. It is worth noting that Daglish et al.’s best-fitting model is actually a third option — the stochastic square-root-of-time rule, which achieves R²=97.12% — but all models that outperform sticky-strike share the same underlying implication: the SPX volatility surface moves with spot, not independently of it.

This is the foundation for the delta correction. When the surface shifts with spot, ∂σ_imp/∂S is consistently negative for fixed-strike options, and the practitioner delta — which assumes ∂σ_imp/∂S = 0 — is wrong in a predictable direction.

What the Numbers Say

Hull and White (2017) estimate E(∂σ_imp/∂S) empirically for S&P 500 options using rolling 36-month windows and find it is well approximated by a quadratic function of the option’s BS delta divided by the product of spot and the square root of time to maturity. This produces a correction formula with three free parameters, estimated once per month:

Δ_MV ≈ Δ_BS + ν × (a·Δ²_BS + b·Δ_BS + c) / (S·√T)

The three coefficients a, b, c are generally stable through time, though Hull and White note extreme parameter shifts during the 2008 credit crisis as the documented exception. They are re-estimated monthly using all strikes and maturities in the prior 36 months, and the results are not sensitive to the choice of window length between 12 and 60 months. One transparency note: Hull and White do not report sub-period performance, so the contribution of the 2008 crisis to the aggregate 25.7% figure is unobservable from the published results. The crisis period produced documented parameter instability; whether the model’s outperformance holds if that period is isolated is a question the paper does not answer. For a desk running tail-risk books, this gap in the published evidence is material.

The out-of-sample test runs from January 2007 to August 2015, covering the 2008 credit crisis. The Gain — Hull and White’s metric for percentage reduction in the sum of squared hedging errors, a variance measure before transaction costs — follows a clear gradient by moneyness: for deep OTM calls (BS delta near 0.1), 42.1%; for delta-0.2 calls, 35.8%; at ATM (delta near 0.5), 27.1%; for deep ITM calls (delta near 0.9), 16.6%. The average across all call strikes is 25.7%. For put options, the average gain is 22.5%, lower because idiosyncratic noise in put implied vol is higher and less of the vol variation is explained by price changes. Hull and White trace this put-call asymmetry to violations of put-call parity in the pre-2009 period; post-2008, the asymmetry narrows.

The gains hold across related index instruments: 23.0% for European-exercise S&P 100 calls (XEO), 16.7% for American-exercise S&P 100 calls (OEX), and 26.5% for DJIA calls. They collapse for individual stocks — 10.3% for calls on Dow components, a statistically negligible 2.5% for puts — and are minimal for interest rate ETFs (1.4%). The correction is an equity index phenomenon, driven by the systematic and stable negative vol-price relationship that characterizes large, liquid indices but is overwhelmed by idiosyncratic noise at the single-stock level.

The Objection: Stochastic Volatility Models Already Solve This

The natural rebuttal is that stochastic volatility models — Heston, SABR — already incorporate the vol-price correlation through the parameter ρ. A trader using SABR-derived delta is in principle computing something close to the minimum variance delta, because the model’s estimated ρ captures exactly E(∂σ/∂S). This is correct in theory. The empirical result is that it fails in practice.

Bakshi, Cao, and Chen (1997) examined three stochastic volatility specifications against S&P 500 options from June 1988 to May 1991 and found that stochastic volatility alone provides the best hedging performance among all models tested — adding jumps or stochastic interest rates does not further improve performance once stochastic vol is included. The model-implied delta with calibrated ρ does reduce hedging error. But the question is not whether SV models beat constant-vol BS; it is whether they match a simple empirical correction.

Hull and White’s abstract states the comparison explicitly: the empirical model outperforms stochastic volatility models “even when the latter are calibrated afresh each day for each option maturity.” The daily calibration of SABR is the deliberately favorable condition for SABR — not a methodological oversight. Hull and White do not test monthly SABR calibration, so the direct comparison is unavailable. What the paper’s own explanation implies, however, is that monthly SABR would likely perform no better. The stated mechanism for SABR’s underperformance is overfitting and model misspecification: “daily recalibration introduces noise into the estimated ρ that the monthly rolling regression avoids.” If overfitting is the cause, reducing calibration frequency would remove noise, potentially shrinking the gap — but the paper’s logic runs in the direction of the daily SABR already being suboptimal relative to a more stable estimator. Monthly SABR may do better or worse; the paper does not resolve this. What can be stated is that SABR with maximum calibration frequency, given every data advantage, still trails the empirical model.

SABR, calibrated daily for every eligible option maturity, requires roughly 87,000 total parameter estimates across the full test period — approximately 40 per trading day on average. The paper’s figure of 78 parameters per day refers to the maximum on days when all 13 tracked maturities pass Hull and White’s data quality filters; on average, roughly 6 to 7 qualifying maturities are available per side per day, not 13. The empirical model, by contrast, estimates three coefficients once per month. It achieves a hedging gain of 24.6% for calls and 19.0% for puts versus the empirical model’s 25.7% and 22.5%. The Newey-West adjusted t-statistics for the difference exceed 8 for all calls and 11 for all puts, each significant at any conventional threshold. The only buckets where SABR leads are deep-in-the-money options — a region of thin volume where Hull and White explicitly note the exception.

Alexander, Rubinov, Kalepky, and Leontsinis (2012) confirm the same direction in a different market: 16.5 years of FTSE 100 options data, where Markov-switching smile-adjusted deltas reduce hedging errors to roughly 50–60% of implied BSM hedging errors on average across all regimes, with substantially greater improvement during volatile periods. The vol-price elasticity is not constant — it is weaker in trending markets and stronger in volatile ones — and the rolling estimation window in Hull and White’s model captures this implicitly.

Why the Gap Persists

The persistence of the practitioner BS delta as the industry standard is not informational. The empirical literature establishing its inferiority spans two decades: Bakshi, Cao, and Chen (1997), Coleman, Kim, Li, and Verma (cited in Hull and White 2017) on S&P 500 options as early as 2001, Crépey (2004), Alexander et al. (2012), and Hull and White (2017). The gap between knowing and implementing — which has persisted across those two decades of cumulative evidence — has three institutional sources.

The practitioner delta is the default output of standard risk systems. Compliance tests, delta limits, and intraday P&L attribution are built against this number. Switching from the partial derivative convention to a minimum variance convention requires changing not just a model but a risk infrastructure, including audit trail requirements and regulatory approval for VaR frameworks.

The correction is index-specific. A mixed book of single-stock and index options captures diluted benefits: individual equity calls show 10.3% gain, individual puts 2.5%. For a desk with material single-stock optionality, the payoff-to-implementation ratio is lower than for a pure index book.

The corrected delta is lower than the practitioner delta for calls. In a rising market, a smaller delta hedge costs less to carry and generates better hedge P&L. In a flat market, the smaller hedge produces lower variance reduction. Managers who evaluate hedging quality on sharp down-days — when delta is clearly insufficient — are not the same managers who evaluate on variance reduction across all trading days. The improvement appears in the variance metric, not the crisis-day metric.

The result is a structural variance cost that accumulates in books using practitioner delta relative to those running smile-aware delta — not a direct extraction of P&L from counterparties with worse hedges, since the underlying’s move dominates any individual rebalance’s realized P&L, but a compounding statistical drag across thousands of daily hedges. Dealers running proprietary smile-aware implementations carry lower realized hedging variance over time; the economic value of that difference is not directly observable from the published evidence but shows up in lower residual risk per unit of notional.

What Would Change This View

Three conditions would reduce or eliminate the minimum variance correction.

The correction depends on a stable negative vol-price correlation for equity indices. If this correlation reverted to zero or turned positive — under a structural regime shift where volatility becomes demand-driven and disconnected from the leverage effect — the term E(∂σ_imp/∂S) would vanish and the practitioner delta would become optimal. The negative vol-price relationship has been stable across decades for large-cap equity indices, but it is not a mathematical law.

If sub-14-day options dominate the book, the correction loses efficacy. Hull and White exclude sub-14-day options from their test and note that including them worsens results because large near-the-money gamma generates P&L variance that a delta correction alone cannot address. The 2008 credit crisis also produced documented extreme parameter shifts in the correction coefficients, a caveat Hull and White flag explicitly. The model is stable under normal conditions but not immune to regime breaks.

If the correction became widely implemented, the systematic pricing asymmetry between buy-side and dealer books would narrow. The current state — that even SABR, which theoretically captures the correction through ρ, underperforms the simple empirical formula — suggests the correction is underexploited even at well-resourced institutions. That underexploitation is what keeps the gap alive.

The Actionable Implication

A PM running vanilla equity index options can compute the minimum variance delta directly from standard risk system outputs. The inputs are the practitioner BS delta (Δ_BS), the practitioner BS vega (ν), current spot (S), and time to maturity (T). The three coefficients must be estimated from historical data, but Hull and White’s published findings establish that estimates derived from any 36-month window of daily option closing data are robust across window-length choices between 12 and 60 months and apply consistently across all strikes and maturities simultaneously.

The correction matters most where it is most frequently ignored: for deep OTM index calls (BS delta near 0.1), the standard convention over-hedges by the amount that generates a 42.1% variance penalty. These are the instruments used by institutional desks as upside participation and by systematic vol strategies as short-gamma positions. The over-hedging of OTM calls and under-hedging of OTM puts is not noise — it is a structural feature of the practitioner convention applied to a vol surface that demonstrably moves with spot.

The 25.7% reduction in sum-of-squared hedging errors for S&P 500 call options, achieved out-of-sample from 2007 to 2015, is not sensitive to parameter choice or window length. It measures the difference between measuring what an option actually does when spot moves and assuming the vol surface is frozen while it moves. The former is a three-number estimate computable from three years of daily option closing prices — matching the rolling window Hull and White used to produce every headline figure in the paper. The latter is the industry default.

Primary sources: Hull, J. and White, A. (2017), “Optimal Delta Hedging for Options,” Journal of Banking and Finance, 82, 180–190. Daglish, T., Hull, J. and Suo, W., “Volatility Surfaces: Theory, Rules of Thumb, and Empirical Evidence,” working paper, University of Toronto. Bakshi, G., Cao, C. and Chen, Z. (1997), “Empirical Performance of Alternative Option Pricing Models,” Journal of Finance, 52(5), 2003–2049. Alexander, C., Rubinov, A., Kalepky, M. and Leontsinis, S. (2012), “Regime-Dependent Smile-Adjusted Delta Hedging,” Journal of Futures Markets, 32(3), 203–229. Crépey, S. (2004), “Delta-Hedging Vega Risk,” Quantitative Finance, 4, 559–579.

Connect with me on LinkedIn, or for video breakdowns of research like this, subscribe on YouTube.

📊 Want Deeper Quantitative Analysis?

This research took a long time of data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

D.E. Shaw, Citadel, and Renaissance are among the largest practitioners of statistical arbitrage, running mean-reversion strategies on the Ornstein-Uhlenbeck framework that has been the field standard since Avellaneda and Lee’s 2010 paper. That framework carries an OLS estimation bias in its core mean-reversion parameter documented in Biometrika since 1954, and the Avellaneda-Lee paper that established the standard does not address it. The dominant explanation for declining statistical arbitrage alpha is crowding: more capital chasing fewer structural mispricings, signal half-life compressed, position correlations rising until forced liquidation triggers cascade losses. That story is right about the facts and incomplete about the mechanism. A second, causally distinct error runs in parallel: OLS estimation of κ in the discretized Ornstein-Uhlenbeck process is finite-sample biased by an order inversely proportional to the calendar span T of the estimation window, not the sample size n. A fund shifting from daily to hourly sampling over the same 60-day window reduces this error by exactly zero. At the Avellaneda and Lee (2010) standard of T₁ = 60 trading days, the Marriott-Pope bias drives calibrated half-lives below their true values, causing funds to trigger exits while the majority of the spread deviation remains open. How large this error is relative to crowding-driven signal decay is not quantifiable from public data. What is established is the mechanism: it exists, it has a known structure, and the fix is available.

The Consensus and the Layer Beneath It

The standard account is empirically grounded. Gatev, Goetzmann and Rouwenhorst (2006) documented average annualized excess returns of up to 11% for the top pairs portfolios over 1962–2002, measured after controlling for bid-ask bounce via a one-day delay but before explicit commissions; the paper states that profits “typically exceed conservative transaction-cost estimates.” Do and Faff (2010) showed, using CRSP data from 1962–2009, that pairs trading profitability was strongest in the period before 1989 and declined from the 1990s onward; a finding widely attributed to their paper across independent academic replications, though the paper’s abstract describes only “the continuing downward trend” without sub-period labels. McLean and Pontiff (2016) showed that across 97 cross-sectional return predictors (a broad sample that includes factor-based strategies but not specifically OU stat arb), portfolio returns were 26% lower out-of-sample (the paper describes this as “an upper bound estimate of data mining effects”) and 58% lower post-publication, with an estimated 32% decline attributable to publication-informed trading. Applying this to OU stat arb is an inference, not a direct measurement.

That account explains erosion of the SIGNAL: the statistical relationship between spread deviation and subsequent convergence. It does not explain errors in how that signal is PROCESSED after entry. A fund that correctly identifies a profitable spread, correctly enters, and exits at the wrong time loses alpha from an implementation failure that is causally separate from crowding, unaffected by how many other funds are in the same trade. These two loss sources have different mechanisms and different remedies. How much each contributes to observed Sharpe compression cannot be resolved from publicly available data.

The Avellaneda-Lee Estimation Architecture

Avellaneda and Lee decompose individual stock returns into a systematic component, explained by sector ETFs or principal components, and an idiosyncratic residual dXᵢ, modeled as:

All three parameters are estimated from a rolling 60-day window (T₁ = 60/252 ≈ 0.238 calendar years). The paper states entry at any residual that deviates by 1.25 standard deviations from equilibrium, and exit when the residual is less than 0.5 standard deviations from equilibrium, a threshold specified uniformly across all stocks. The tradeability filter requires κᵢ > 252/30 = 8.4 per year (half-life under 30 trading days).

From κ̂ᵢ flows every quantity that matters operationally: the S-score

The position size, and the tradeability filter. If κ̂ is materially wrong, all of them are wrong simultaneously.

The paper covers 1997–2007. PCA-based strategies averaged Sharpe 1.44 over that full window, a figure that already incorporates the weak later years; during 2003–2007 alone, the sub-period Sharpe was only 0.9. The paper describes pre-2003 performance as “much stronger” without reporting the sub-period separately. Working backward from the stated averages: if the full 11-year average is 1.44 and the 2003–2007 (5-year) average is 0.9, the implied 1997–2002 (6-year) average is approximately 1.89, meaning the within-sample drop was from roughly 1.89 to 0.9. Post-2010 performance of the vanilla OLS implementation is unobservable from public primary sources.

The T⁻¹ Bias: Mechanism and Lineage

The finite-sample bias in AR coefficient estimation was first formally analyzed by Hurwicz (1950) in the Cowles Commission Monograph on dynamic economic models. Marriott and Pope (1954) derived the first analytical approximation in Biometrika: for an AR(1) with intercept, the OLS estimate of the autoregressive coefficient â₁ is biased by approximately −(1 + 3a₁)/n. That discrete-time result sat in the econometrics literature for 55 years before Tang and Chen (2009) adapted it to the continuous-time OU process in the Journal of Econometrics, establishing that the bias in κ̂ is of order T⁻¹ (total calendar span), not n⁻¹ (sample size). Yu (2012) refined the formula with a nonlinear correction, identifying that the Marriott-Pope approximation “does not work satisfactorily when the speed of mean reversion is slow”: the near-unit-root regime, because in that limit, Yu showed, “the true bias has an interesting curvature and goes to zero when the mean reversion parameter is closer to zero.” This correction matters for the worked examples below.

The T⁻¹ vs. n⁻¹ result is the operationally decisive finding. A fund sampling daily over 60 days (n = 60, T = 0.238 years) faces bias of order 1/T ≈ 4.2 per year. The same fund sampling hourly (n = 1,260, T unchanged) faces the same bias. Only extending the window to 120 days halves it.

The Marriott-Pope Cascade: What the First-Order Approximation Shows

When practitioners estimate κ by regressing Xₜ₊₁ on Xₜ and a constant to obtain â₁, then computing κ̂ = −ln(â₁)/Δt, the Marriott-Pope bias in â₁ (toward zero) translates into an upward bias in κ̂:

Mean-reversion appears faster than it is. The estimated half-life

falls below the true half-life.

The calculations below use the first-order Marriott-Pope approximation, Bias(â₁) ≈ −(1+3a₁)/n. This approximation is more reliable away from unit root and deteriorates as a₁ approaches 1. In the near-unit-root case (the κ = 3.8 example, where a₁ ≈ 0.985), Yu (2012) shows the true bias is actually smaller than Marriott-Pope predicts, because the true bias in κ approaches zero as κ approaches zero. That example should be read as an upper bound on the bias magnitude in that regime, not a point estimate. The κ = 15 example, where a₁ ≈ 0.942, is less affected by this limitation and is the more reliable illustration.

For a pair with true κ = 15/year (true half-life ≈ 11.6 trading days):

• Daily AR coefficient: a₁ = e^{−15/252} ≈ 0.942

• Marriott-Pope bias: −(1 + 3 × 0.942)/60 ≈ −0.064; estimated â₁ ≈ 0.878

• κ̂ ≈ 33.0/year (120% above true κ under this approximation)

• Estimated half-life ≈ 5.3 trading days (vs. true 11.6)

For a pair with true κ = 3.8/year (true half-life ≈ 46 trading days), in the near-unit-root regime where Marriott-Pope overstates the bias:

• a₁ = e^{−3.8/252} ≈ 0.985; estimated â₁ ≈ 0.919 under Marriott-Pope

• κ̂ ≈ 21.3/year under this approximation; this pair would pass the κ > 8.4 filter

• The 460% overstatement derived from Marriott-Pope is an upper bound; the true magnitude requires Yu’s numerical formula

What the two cases illustrate mechanistically: a pair that should fail the tradeability filter (true half-life 46 days, which is outside the intended < 30-day screen) can pass it because biased κ estimation makes it look fast. And a pair that is truly fast-reverting gets assigned an estimated half-life that is a fraction of the true value, causing the exit signal to fire before most of the convergence has occurred.

The fraction of reversion remaining at the biased exit point (first-order approximation):

For κ = 15, κ̂ = 33 (the more reliable example): 2^{−15/33} ≈ 0.73 under this approximation. This illustrates that if the approximation holds, the exit signal fires while roughly 73% of the spread deviation remains open.

These are derived quantities from a first-order formula applied to chosen illustrative parameters, not measurements from any live book. Their value is in showing that the mechanism, if present, would not be a small second-order effect; even under a conservative approximation, the bias ratio κ̂/κ ≈ 2 is large enough to matter. Whether it matters as much as crowding in any specific implementation is unresolvable from public data.

The Standard Error Reinforces the Same Conclusion

For κ = 15/year with 60 daily observations, the standard error of κ̂ from the AR(1) regression via delta method is approximately SE(κ̂) ≈ 252 × √((1−a₁²)/n) / a₁ ≈ 11.9 per year.

Centered on the true κ = 15, the 95% interval for what an analyst would observe spans roughly [−8, 38] per year, essentially uninformative about whether the true half-life is 5 days or 60 days. Centered on the observable κ̂ ≈ 33 (the only value an analyst actually has in production), the interval spans roughly [10, 56] per year, still wide enough to be consistent with the true κ ranging from near-zero to very fast reversion. Both framings point to the same conclusion: the 60-day window cannot reliably identify the parameter that governs every downstream exit decision.

Crowding Is a Bias Amplifier, Not the Primary Source

When multiple funds hold identical long-short positions, their collective flow physically accelerates convergence, raising the realized κ in the estimation window. A fund running 60-day OLS calibrates to this crowding-inflated speed and sizes up accordingly.

When any fund begins to liquidate (the “Unwind Hypothesis” of Khandani and Lo, NBER Working Paper 14465 (2008)), the coordinated flow stops. Remaining funds are positioned to a κ that only existed under crowded conditions.

Khandani and Lo found in the NBER paper that “the expected return of a simple mean-reversion strategy increased monotonically with the holding period during this time, i.e., those marketmakers that were able to hold their positions longer received higher premiums.” Two limitations apply when invoking this as evidence for the bias-exit thesis: first, the strategy they simulated is the Lehmann (1990)/Lo-MacKinlay (1990) daily contrarian, not the Avellaneda-Lee OU approach; the holding-period premium is consistent with the early-exit mechanism but is not a test of it. Second, Khandani and Lo explicitly state that “the hypotheses advanced in this paper are speculative, tentative, and based solely on indirect evidence”; they had no access to fund-level position data. The inference chain from that paper to the bias-exit mechanism requires multiple steps, each of which adds uncertainty.

Decay and Capacity: What the Mechanism and What the Unknowns Are

The underlying signal (idiosyncratic mean-reversion after factor neutralization) is subject to publication-informed decay. Applying the McLean-Pontiff framework by inference, with the caveat that their study covers cross-sectional predictors broadly rather than OU stat arb specifically, the GGR publication in 2006 likely accelerated capital flows into the strategy. The within-sample degradation Avellaneda and Lee document is consistent with that trend: from a pre-2003 implied Sharpe of roughly 1.89 to 0.9 in 2003–2007. The signal still exists; McLean and Pontiff found that post-publication returns are more durable in high-idiosyncratic-risk, low-liquidity securities.

The bias-correction is not a published trading signal; it is a calibration fix to an existing implementation. It is not subject to McLean-Pontiff decay because knowing about it does not let competing capital trade against it. Whether any specific fund has implemented the Tang-Chen bootstrap or the Yu (2012) analytical correction in their production κ calibration is unknown from public sources. The claim that PhD-staffed stat arb desks are unaware of a bias documented since 1954 is not a credible prior; the more honest question is whether they have specifically applied the correction to their 60-day OU estimation pipeline, which is a different question with no public answer.

The practical point is not “this alpha is sitting on the table uncaptured.” The practical point is: if you are running 60-day OLS κ estimation, the diagnostic test of applying Tang-Chen correction and comparing κ̂ to κ̂_corrected will tell you whether your implementation has this problem and how severe it is in your specific universe. That is the actionable step. Whether the result turns out to matter a lot or a little depends on your specific pair universe, and that test can only be run against your own production data.

What Would Change This View

Three empirical findings would substantially weaken the thesis:

First, an intra-trade P&L decomposition showing that Avellaneda-Lee alpha concentrates in the early portion of the hold rather than later would suggest premature exit is not the operational failure mode. This data is not publicly available.

Second, evidence that implementations already using bias-corrected κ (via Tang-Chen bootstrap, Yu analytical formula, or longer windows) show no systematic improvement in exit timing relative to OLS 60-day implementations would suggest the mechanism, while theoretically present, is not materially significant in practice.

Third, evidence that the variance of κ̂ at n = 60 so overwhelmingly dominates the bias that the directional correction is noise-overwhelmed would reduce the prescription to “extend the window regardless.” The wide confidence interval already documented is consistent with this possibility.

Interventions Worth Investigating

These are not “here is the discovered alpha” recommendations. They are diagnostics that follow from the documented mechanism and are worth running against any implementation currently using 60-day OLS κ estimation:

1. Compare bias-corrected versus raw κ̂. Run the Tang and Chen (2009) parametric bootstrap on your existing calibration: simulate 200+ paths from the fitted OU model, re-estimate κ on each path, subtract the estimated bias. Compare κ̂_corrected to κ̂. The distribution of the ratio κ̂/κ̂_corrected across your universe, over time, will tell you whether the bias is large enough to affect your filter and exit timing materially.

2. Test the 120-day window against the 60-day window. Extend T₁ from 60 to 120 days in a shadow book and compare exit timing and P&L per trade against your live implementation. The O(T⁻¹) scaling predicts a halving of the bias-related error; whether that translates to improved P&L in your universe is an empirical question only your data can answer.

3. Audit the tradeability filter. At n = 60, the Marriott-Pope formula predicts that pairs near unit root (true κ ≈ 2–5 per year) will systematically produce κ̂ well above 8.4. If your universe contains slow-reverting residuals masquerading as fast reverters, identifying and re-screening them may reduce capital deployed in positions where the holding period assumption is violated.

The capacity constraint on any of these interventions is the same as the underlying stat arb book: not estimable from public sources. The correction itself has no additional market impact cost.

📊 Want Deeper Quantitative Analysis?

This research took a very long time of data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

Primary sources: Avellaneda and Lee (2010), Quantitative Finance 10(7): 761–782 | Gatev, Goetzmann, Rouwenhorst (2006), Review of Financial Studies 19(3): 797–827 | McLean and Pontiff (2016), Journal of Finance 71(1): 5–32 | Do and Faff (2010), Financial Analysts Journal 66(4): 83–95 | Hurwicz (1950), Chapter XV in Koopmans ed., Statistical Inference in Dynamic Economic Models, Cowles Monograph 10 | Marriott and Pope (1954), Biometrika 41(3–4): 390–402 | Tang and Chen (2009), Journal of Econometrics 149(1): 65–81 | Yu (2012), Journal of Econometrics 169(1): 114–122 | Khandani and Lo (2008), NBER Working Paper 14465

Follow for more quantitative finance research: YouTube: The Mathematical Trader | LinkedIn | Patreon

For a decade, the clearest method of generating alpha from SpaceX was not owning SpaceX. It was owning the access-constrained proxies that held it by necessity, at premiums that reflected the structural gap between demand and supply. S3 Partners confirmed the mechanism in the week following the June 12 IPO: “SpaceX’s record IPO removes the central constraint of the space trade — for years it meant buying public proxies because the category leader wasn’t listed.” With Phase 1 fully captured and Phase 2 decaying on a disclosed calendar through December 9, 2026, the one alpha channel still open requires correctly disaggregating why space stocks sold off on IPO day, but that disaggregation only holds if you understand what the ASTS bear case actually rests on, which is not the satellite count.*

🎬 Prefer to watch rather than read? A NotebookLM-generated video overview of this article is available here: Watch the video overview → Full analysis, citations, and data remain in the article below.

Phase 1: The Proxy Access Alpha (2008–2026, Captured)

SpaceX’s private status imposed an access constraint that most institutional capital could not cross. Investors who needed exposure held it through vehicles that could reach the secondary market or had carried the position for years, while those vehicles traded at premiums reflecting the optionality of owning the category leader before it listed. The figures below are drawn from the best available primary sources: state regulatory filings, fund investor communications, and contemporaneous press releases, but none are audited returns, and that limitation applies to the entire section.

Google’s approximately $900 million investment in SpaceX’s January 2015 round, an amount consistently estimated across secondary sources including Bloomberg and Fortune but not disclosed in Alphabet’s SEC filings, produced a stake that a 2026 Alaska state filing reported by Bloomberg placed at 6.11% of SpaceX at end-2025, since diluted to approximately 5% following the February 2026 xAI merger. At the $1.75 trillion IPO price, that range represents a paper gain of approximately 97x to 120x on the estimated cost basis, with the cost basis itself unverifiable from any Alphabet filing.

Founders Fund invested approximately $20.4 million in SpaceX’s Series C in August 2008, confirmed in a contemporaneous SpaceNews press release. The firm last appeared in Alaska’s state shareholder disclosure at a 5.76% stake as of December 2023, per Bloomberg. The stake has since fallen below the 5% reporting threshold, meaning Founders Fund has sold down some portion and its current holdings are unverifiable from public records. Founders Fund partner Scott Nolan told Axios in May 2026 the remaining stake “could be worth more than $60 billion.”

Scottish Mortgage, Baillie Gifford’s flagship UK investment trust, invested GBP £151m (approximately $200m at time of purchase) in SpaceX between December 2018 and August 2021. Per its pre-IPO briefing note of May 2026 (an investor communication, not an audited filing) the stake carried a fair value of GBP £2.98bn as of March 31, 2026, a stated 19.7x return on invested capital. The same document describes SpaceX as “our biggest contributor to returns over 1, 3 and 5 years.”

For funds without direct secondary market access, DXYZ (Destiny Tech100) offered a publicly listed vehicle with SpaceX as its largest holding at 16.2% of portfolio. DXYZ reported a net asset value of $19.97 per share as of December 31, 2025, per its Business Wire press release, while trading at a substantial premium to that figure in the months preceding the IPO, reflecting the pure optionality value attached to pre-IPO SpaceX exposure.

Decay: Complete as of June 12, 2026. Forge Global states it no longer tracks SpaceX. Any fund still holding proxy vehicles at a premium to NAV for SpaceX exposure is paying the access premium for a structural constraint that no longer exists.

Capacity: Limited pre-IPO to the secondary market’s effective throughput, with individual institutional tender-offer participation capped by SpaceX’s own structure. The edge was access-constrained rather than market-impact-constrained.

Phase 2: The Float Mechanics Trade (June 12 to December 9, 2026, Decaying)

SpaceX listed with an initial float of approximately 555.6 million shares, expanded to approximately 638.9 million after underwriters exercised the full greenshoe on June 15, producing a public float of approximately 4.9% of 13.08 billion total shares. This float met an estimated $35 to $50 billion in forced MSCI passive buying concentrated into roughly $70 billion in tradeable shares. SPCX ran from the $135 IPO price to an intraday peak of $225.64 on June 16, a 67% move driven by supply-demand mechanics, not by any new information about the businesses. It now trades at approximately $154.

The staggered lock-up schedule is disclosed in the 424B4 prospectus: a first tranche releasing approximately 20% of the 180-day lock-up block on the second full trading day after August 6 earnings; staggered 7% tranches at days 70, 90, 105, 120, and 135 post-IPO; a Q3 earnings-triggered 28% release; full 180-day expiry on December 9, 2026. Musk’s approximately 6.4 billion Class B shares unlock June 12, 2027.

Decay: Mechanical and calendar-driven. S&P 500 exclusion is structural: the index requires GAAP profitability in both the most recent quarter and cumulatively over four quarters. SpaceX posted a $4.94 billion GAAP net loss in 2025 and approximately $4.3 billion in Q1 2026, a range reported as $4.27B to $4.30B across credible sources reading the same prospectus. The GAAP losses are driven by xAI’s $6.355 billion operating loss on $3.20 billion in 2025 revenue. The vast S&P 500-linked passive capital base, excluded from buying SPCX on inclusion criteria, remains a structural buyer waiting until SpaceX posts four consecutive GAAP-profitable quarters, not realistic on current trajectory before late 2027.

Capacity: Approximately $2B AUM before position sizing creates material price impact against the current float. Inaccessible to larger funds without moving the market.

Phase 3: The Post-Proxy Disaggregation Trade

S3 Partners identified a clean divergence in how capital moved through the space sector around the IPO. Short interest in Rocket Lab declined 22% year-to-date, while short interest in AST SpaceMobile built 41% year-to-date to approximately 54 million shares. S3 estimates only 22% of ASTS short interest is convertible-bond arbitrage, meaning approximately 78% is directional. The firm’s characterization: “Rocket Lab and Viasat, the cash-generating operators, ran the other way: shorts covered both over the past month and YTD. At the genuinely directional shorts, AST SpaceMobile’s +41% build and Virgin Galactic’s +86%, the short side outran the long.”

The IPO-day rotation sold all space stocks simultaneously. Rocket Lab fell approximately 10.8%. Analysts called it capital rotation. That characterization is correct for RKLB and structurally incomplete for ASTS, where the rotation and a genuine fundamental repricing arrived simultaneously, priced as a single event when they are not the same trade.

Why the two names have different exposures to SpaceX

Rocket Lab is a launch services and space systems provider. Its Q1 2026 SEC 8-K filed May 7, 2026, reports record quarterly revenue of $200.3M (+63.5% year-over-year), record backlog of $2.22B (+108% year-over-year), and Q2 2026 guidance of $225M to $240M against a consensus of $207.5M. Its backlog includes an $816M Space Development Agency Tracking Layer Tranche 3 contract for missile warning and defense satellites and a $515M SDA Transport Layer-Beta Tranche 2 program, totaling more than $1.3 billion in contracted SDA work awarded competitively from a field that did not include SpaceX on TRKT3 specifically. SpaceX competed in the same SDA ecosystem via the $2.29 billion SDN Backbone award and a $4.16 billion Air Moving Target Indicator contract in May 2026, both representing a separate and larger program enclosure. The medium-term risk is real: as SatNews reported, the week of those SpaceX awards was “the seven days in which American military space lost its remaining alternatives,” and RKLB’s SDA pipeline beyond its current contracted work faces a genuine consolidation risk. What is not at risk from SpaceX competition today is the existing contracted backlog of $1.3B in SDA work already awarded. RKLB joined the Nasdaq-100 on June 22, 2026, creating structural passive buying demand. Neutron, targeting a Q4 2026 debut at $50 to $55M per mission versus Electron’s approximately $8M, represents a step-change in unit economics if it delivers.

The pair trade is not a bet that RKLB is immune to SpaceX competition in government space. It is a bet that the existing contracted backlog, Nasdaq-100 inclusion demand, and Neutron optionality constitute a fundamentally different risk profile from ASTS, and that the IPO-day rotation priced them identically.

The ASTS Bear Case: What It Actually Rests On

The satellite count argument (9 operational BlueBirds today versus Starlink’s 650-plus DTC satellites) is true but insufficient as a thesis anchor on its own, because ASTS is assembling satellites at 6 per month and announced BB11–13 launching in the first half of August from Cape Canaveral, per a BusinessWire release on June 23. At the August 6 catalyst date, ASTS will likely have 12 operational BlueBirds, a ratio of roughly 54:1 against Starlink, better than the current 72:1, but directionally the same. By year-end 2026 at the targeted 45 satellites, the ratio is approximately 14:1. A directional short seller reading the satellite count will immediately ask whether ASTS’s trajectory invalidates the thesis. It doesn’t, but the reason why is not the satellite count.

The correct anchor is ASTS’s 10-K revenue recognition policy, which states verbatim: “Revenue for SpaceMobile Service will be recognized over the life of the contract, beginning when AST provides MNOs access to its satellite network. To date, the Company has not recognized any revenues from its SpaceMobile Service.” The $1.2 billion in contracted MNO revenue commitments does not recognize on satellite deployment milestones. It does not recognize on carrier infrastructure investment by partners. It recognizes when ASTS provides network access, meaning commercial service launch, which the company targets for H1 2027 and requires reaching a minimum constellation density of 25-plus satellites for limited non-continuous service. The 45-satellite year-end target is necessary for that commercial launch, not sufficient in itself for revenue recognition.

This means the Q1 2026 revenue miss requires a more precise interpretation than its surface appearance suggests. The $14.7 million in Q1 came from gateway equipment deliveries and government contract milestone achievements, not from SpaceMobile Service. ASTS’s own Q1 press release described the result as “consistent with plans for quarterly revenue ramp.” The miss was against Wall Street consensus, not ASTS’s internal plan. Q2 will test something more specific: whether gateway deliveries and government milestones can accelerate enough to reach the $40 to $50 million quarterly run-rate implied by the $150 to $200 million full-year guidance, with the additional variable of whether initial commercial service activations in targeted markets generate any SpaceMobile Service revenue recognition in H2 2026.

The structural bear case: even at 45 satellites and commercial service launch in H1 2027, Starlink DTC will have 18 months of commercial data with T-Mobile, multiple carrier partnerships, and a proved product. The moment ASTS’s $1.2B in contracted MNO revenue starts recognizing, those same MNOs face a choice between a committed supplier and a proven incumbent. The satellite trajectory accelerating through year-end is a necessary operational milestone, not a revenue event. These are different things, and conflating them is how the bull case overstates the near-term.

As of December 31, 2025, ASTS carried $227 million in deferred contract liabilities (cash received from partners including a $175 million prepayment from stc Group), sitting on the balance sheet because the performance obligation (network access) has not yet been satisfied. That $227M represents commercial demand. It also represents revenue that won’t hit the income statement until ASTS crosses the commercial service threshold in 2027.

ASTS Q1 Earnings in Context

ASTS’s Q1 2026 results showed revenue of $14.7M against estimates of $37.5M to $39M, with net loss of $191M against an expected $86.8M loss, EPS of -$0.66 against a -$0.21 to -$0.24 estimate, and total operating expenses of $164.1M. The 60%-plus miss against consensus was driven by the inherently lumpy nature of gateway delivery timing and government milestone achievements, not by SpaceMobile Service revenue failing to materialize; that revenue doesn’t start until commercial launch. The company targets full-year 2026 revenue of $150 to $200 million, implying a significant ramp in Q2 through Q4 from gateway sales and milestones. That ramp is the August 6 test, not whether the 12-satellite constellation has commercialized.

The three-scenario expected-value model, position sizing at 10% ADV participation, stop-loss triggers, borrow mechanics, and the specific falsification condition that forces an immediate exit before August 6 are in the full institutional note on Patreon.

The Morningstar Reference: What the $62 Includes

Morningstar’s $62 fair value estimate for SPCX was published in the context of the Cursor acquisition analysis, revised down from $63 at IPO to $62 after the $60 billion all-stock deal. The estimate accounts for Cursor dilution. The market price of approximately $154 also reflects post-Cursor trading. Both figures are post-Cursor, making the 2.5x premium ratio consistent on the same basis. Morningstar’s methodology yields $40 per share for core space and connectivity operations, $16.50 probability-weighted for the AI segment, and $5.51 net cash less debt, totaling $62.51 before rounding. The ratio of approximately 2.5x at current prices reflects the gap between that sum-of-parts math and the market’s implied valuation of the Starship and orbital computing optionality.

The Strongest Objection

The investor community Juxtaposed on TipRanks published the opposite position as of June 21: bullish on ASTS, hold on RKLB. The ASTS bull case: $3.5 billion in cash, cash equivalents, and restricted cash as of Q1 2026, per the ASTS 10-Q, against $2.97 billion in long-term debt, plus $1.2 billion in contracted revenue commitments from AT&T, Verizon, Vodafone, Rakuten, and others, and BB11–13 launching in early August with production already through BB37. If carriers treat ASTS and Starlink DTC as complementary, and if ASTS reaches commercial launch in H1 2027 as planned, the contracted revenue begins recognizing and the deferred $227M starts flowing.

There is also a material valuation concern on the long side. RKLB trades at approximately 45x forward price-to-sales against SpaceX at roughly 30x, despite RKLB’s substantially lower revenue base, per Seeking Alpha’s post-IPO analysis. On this metric, RKLB is priced more expensively than the category leader. RKLB’s SVP and General Counsel disposed of $9.5M in shares on June 18, 2026, per SEC Form 4 disclosures. And RKLB’s $3.0B at-the-market equity offering creates ongoing dilution. Separately, SpaceX’s growing role in SDA military space, specifically the SDN Backbone and AMTI wins in May 2026, represents a medium-term encroachment on RKLB’s government revenue pipeline beyond the current contracted backlog.

The pair trade survives this objection only if RKLB’s Nasdaq-100 inclusion and government contract backlog provide structural support against its valuation premium and dilution headwind, while ASTS misses its Q2 gateway and milestone run-rate target by enough to cast doubt on full-year guidance. If ASTS demonstrates a clean Q2 acceleration to $40M-plus in revenue driven by gateway deliveries, the timing-miss narrative is validated and the directional short fails.

Decay and Capacity

Long RKLB: Decays if Neutron’s Q4 2026 debut fails or is delayed; if Q2 earnings miss the backlog-implied revenue trajectory; or if RKLB’s SDA pipeline beyond current contracts erodes as SpaceX consolidates military space. The June 18 insider disposal is a monitoring signal. Approximate capacity: at standard 10% participation over five trading days against RKLB’s approximate 20-day average daily volume of $350M to $450M, a position of roughly $175M to $225M stays below standard single-name impact thresholds. These are estimates based on stated participation assumptions, not hard figures; actual market impact depends on execution strategy and prevailing liquidity.

Short ASTS: Decays if ASTS Q2 2026 revenue accelerates materially toward the full-year guidance run-rate, demonstrating the Q1 miss was execution timing rather than structural failure. Per S3 Partners via Yahoo Finance, approximately 30M shares are available to borrow with borrow costs below the squeeze threshold at last report. At the same 10% participation rate against ASTS’s approximate 20-day ADV of $100M to $150M, a position of roughly $50M to $75M is the practical impact threshold; above that level, the short begins moving the price. Total combined capacity for the pair at these participation assumptions: approximately $225M to $300M. Larger positions require longer execution windows and higher cost.

Why the edge persists: The IPO-day rotation was mechanical. Capital exited all space stocks without fundamental differentiation. The RKLB/ASTS fundamental divergence (a $2.22B defense-anchored backlog against a 60%-plus consensus revenue miss, with ASTS revenue recognition structurally deferred to 2027) has been visible in public filings since May 2026. The market priced both names identically on June 12. The S3 short interest data shows that informed directional positioning had already diverged before the IPO. The question is whether August confirms or invalidates that divergence.

What Would Change This View

Three conditions flip the thesis. ASTS Q2 2026 revenue materially beats the $40M to $50M quarterly run-rate implied by full-year guidance, demonstrating that gateway deliveries and government milestones can accelerate beyond Q1’s rate. Starlink DTC adoption stalls from spectrum interference challenges or carrier resistance to the T-Mobile exclusivity structure. Neutron’s Q4 2026 debut fails or is delayed, removing the primary RKLB re-rating catalyst.

A fourth condition that would change the broader three-phase framework: evidence that ASTS’s MNO revenue recognition begins earlier than H1 2027 because network access is provided to carriers for limited commercial use before the full 45-satellite constellation is reached. The 10-K language (”beginning when AST provides MNOs access to its satellite network”) is not satellite-count specific; it triggers on network access provision. If ASTS activates limited commercial service in one or two markets in late 2026 with fewer than 45 satellites, some portion of the $1.2B in contracted revenue begins recognizing sooner than the H1 2027 base case.

The Actionable Implication

Phase 1 alpha is fully captured. It is not replicable. Phase 2 decays mechanically through December 9, 2026, accessible only to funds below approximately $2B AUM. Neither has residual alpha available to a reader who was not already positioned.

Phase 3 is the only channel that does not require timing the float mechanics or taking direct SPCX exposure at a 2.5x fundamental premium to Morningstar’s $62 sum-of-parts estimate. Its combined capacity of approximately $225M to $300M at standard participation rates keeps it from being fully arbitraged by platforms that would otherwise close it immediately, and that ceiling is the structural reason the divergence persists.

The August catalyst is specific: RKLB and ASTS both report earnings within days of the first SPCX lock-up tranche release. Three information streams become simultaneously observable: RKLB Q2 versus its $225M to $240M guidance, ASTS Q2 versus its $40M to $50M implied quarterly run-rate, and the initial insider selling decision from SPCX insiders who for the first time can reveal their private price discovery. Each is analytically distinct. Together, they constitute the most information-dense calendar event between now and December 9.

The pair trade is long a launch company with a record $2.22B contracted government backlog, Nasdaq-100 inclusion demand, and a Q4 Neutron launch option, against short a direct-to-device satellite operator whose $1.2B in contracted MNO revenue is structurally deferred to 2027, whose Q1 revenue was consistent with its own plan but 60%-plus below Wall Street consensus, and whose 12 operational satellites at the August window face a 54:1 disadvantage against Starlink DTC. It is not a bet on valuation convergence. It is a bet on fundamental divergence that IPO-day rotation priced as the same position.

📊 Want Deeper Quantitative Analysis?

This research took a very long time of data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

Full institutional note — three-scenario EV model, stop-loss triggers, borrow mechanics, capacity ceiling, and all primary SEC sources: Patreon.

Primary sources: Rocket Lab Q1 2026 SEC 8-K (May 7, 2026); AST SpaceMobile 10-K FY2025 (revenue recognition policy); ASTS Q1 2026 press release via SEC (March 31, 2026 quarter); ASTS BB11-13 launch announcement, BusinessWire (June 23, 2026); ASTS Q1 2026 earnings, Yahoo Finance; S3 Partners post-IPO space sector analysis, Yahoo Finance; SpaceX prospectus xAI financials, Yahoo Finance; Rocket Lab SDA Tranche 3 contract, SEC press release (December 19, 2025); SpaceX SDN Backbone $2.29B contract, SpaceNews (May 2026); SatNews military space consolidation analysis (June 2026); Morningstar $62 SPCX fair value, post-Cursor (June 14, 2026); Scottish Mortgage pre-IPO briefing note, Baillie Gifford; Alaska state filing on Alphabet stake, Bloomberg; Founders Fund 2008 investment, SpaceNews; Scott Nolan interview, Axios; S&P 500 exclusion, Yahoo Finance; SpotGamma index flow analysis. Investor letters and fund communications are marketing materials, not audited statements, and are attributed accordingly. This article does not constitute investment advice. The author holds no position in any securities mentioned.

Follow the research: YouTube — The Mathematical Trader · LinkedIn · Patreon

The Hill estimator’s cross-sectional equity alpha — 5.4% annually per Kelly and Jiang (2014, Review of Financial Studies) — is dead in the form it was published. Hou, Xue, and Zhang (2020, Review of Financial Studies) find t-statistics of 0.57–1.13 in full-sample replication, down from the original 2.0–2.15, consistent with the decay McLean and Pontiff (2016, Journal of Finance) document when no structural barrier protects an anomaly from publication-informed trading. The same estimator’s application to the options market — anchoring the volatility surface’s power-law continuation from near-money strikes to extreme ones — has not been arbitraged away, because the barrier here is not analytical effort but a hedgeability constraint that dealer balance sheets and the Basel III Fundamental Review of the Trading Book have jointly locked in place.

The Consensus and Its Omission

The Kelly-Jiang result earned its reputation. Applying the Hill formula

ξ̂ = (1/k) ∑ log(Xᵢ / Xₖ₊₁)

to the cross-section of firm-level daily return crashes each month extracts a time-varying common tail factor λₜ. The paper showed that a one-standard-deviation increase in λₜ predicts 4.5% excess market returns over the following year, and that stocks in the top tail-beta decile earn 5.4% more annual three-factor alpha than stocks in the bottom decile. These numbers appeared in 2013–2014 and were widely read.

What the consensus omits is the replication record. McLean and Pontiff (2016, Journal of Finance) study 97 anomalies and find that portfolio returns are 26% lower out-of-sample and 58% lower post-publication — of which 32 percentage points (58% minus 26%) reflect publication-informed trading and 26 percentage points represent an upper bound on data-mining effects in the original studies. Chen, Lopez-Lira, and Zimmermann (2022, arXiv:2212.10317) find a similar pattern across a broader sample: approximately 50% of predictability remains after the original sample periods, a result that holds for risk-based and theory-motivated research categories alongside purely data-driven ones. This is the correct baseline expectation when evaluating a 2014 equity factor paper in 2026.

The Replication Failure, Precisely Quantified

Hou, Xue, and Zhang (2020, Review of Financial Studies, vol. 33, pp. 2019–2133) replicate 452 anomalies using NYSE breakpoints and value-weighted returns. This methodology is standard for avoiding microcap-driven results: when NYSE breakpoints set the decile boundaries, microcap stocks fall into the same deciles by price range but at minimal portfolio weight, preventing a handful of illiquid small-caps from driving the reported alpha.

For the Kelly-Jiang tail risk anomaly, the replication result is explicit:

“The high-minus-low tail risk (Tail) deciles earn on average 0.11%, 0.15%, and 0.19% per month (t = 0.57, 0.79, and 1.13) at the 1-, 6-, and 12-month horizons, respectively. These estimates are lower than 0.36% (t = 2) at the 1-month and 0.35% (t = 2.15) at the 12-month horizon reported in Kelly and Jiang (2014).”

A drop from t ≈ 2.0–2.15 to t ≈ 0.57–1.13 is a collapse, not a haircut. The alpha is statistically indistinguishable from zero under NYSE breakpoints. Imposing |t| ≥ 2.78 across the full HXZ library of 452 anomalies pushes the failure rate to 82.1%.

The mechanism of decay is structural. The original alpha was concentrated in microcap stocks — where NYSE-Amex-NASDAQ breakpoints assign large portfolio weights to small companies that real-money managers cannot hold at scale. Once the factor construction was public, any quant team with CRSP access could implement long-short exposure in high-tail-beta names. There was no proprietary data, no specialized execution infrastructure, no minimum fund size that made replication impossible. Rapid crowding was the predictable result.

What Survives: Relative Mispricing of Deep OTM Options

The evidence for the surviving options-market application comes primarily from Taleb, Yarckin, Mann, Delic, and Spitznagel (2019, revised March 2023, arXiv:1908.02347). The conflict of interest in this source must be named directly: all five authors are principals or employees of Universa Investments, the $20B fund that commercially runs the exact strategy described in the paper. This is a fund whitepaper with academic formatting, not independent academic research. A reader should weight it accordingly.

That said, the mechanism deserves examination independently of who published it. The underlying mathematical claim — that a power-law distribution implies a specific relative price relationship between options at different strikes — rests on established results in extreme value theory that predate and are independent of Universa. The testable empirical claim (that market prices at extreme strikes fall below the power-law continuation from near-money anchors) is falsifiable by anyone with access to an options data terminal. And the equity cross-section result the same authors might cite to motivate their edge was destroyed by independent replication — which actually shows the replication community does correct inflated claims in this space. The options-market mechanism is worth examining on its own terms, with the source’s incentive structure held in view.

Taleb et al. are explicit about scope: “our approach isn’t about absolute mispricing of tail options, but relative to a given strike closer to the money.” The framework uses the Hill-estimated tail index α as the sole parameter for computing option prices beyond any observable anchor strike. Once the return distribution enters the power-law regime — past the “Karamata constant” where the slowly-varying function L(x) stabilizes — relative put prices follow approximately:

P(K₂) / P(K₁) ≈ [(K₂ − S₀) / (K₁ − S₀)]^(1−α)

The exponent is (1 − α). For α ≈ 2.75, this equals −1.75, meaning put prices should decay as (strike distance)^(−1.75) as you move deeper out of the money. Stochastic volatility models (Heston, SABR, local vol) calibrated to the near-money smile extrapolate to extreme strikes with options prices that decay faster than this power law — they imply an effectively higher α (lighter tail) at delta-5 and below than the physically calibrated value.

Taleb et al.’s Figure 3 demonstrates this for the December 31, 2018 S&P 500 settlement: using α = 2.75 and a near-money anchor, the power-law formula produces put prices above market prices at extreme strikes. Deep OTM puts at delta-5 and below trade cheaper than the correct power-law continuation from near-money options would set.

The aggregation objection. The tail index estimates in Gabaix, Gopikrishnan, Plerou, and Stanley (2003, Physica A) — α ≈ 2.70 ± 0.10 (negative tail) and α ≈ 2.96 ± 0.09 (positive tail) — are for individual CRSP stocks binned by market capitalization, not for the S&P 500 index. A knowledgeable reader will immediately flag that diversification should push the index tail index upward relative to constituents: idiosyncratic crashes wash out, and the portfolio should have lighter tails than its components. If the true S&P 500 index α is 3.2 rather than 2.75, the power-law continuation price is lower and the gap between the formula and market prices narrows or disappears.

The empirical answer is that the expected direction does not materialize. Direct Hill estimation applied to the S&P 500 index returns yields α in the range of approximately 2.5–3.0 — similar to or below the individual stock estimates — because the tails of equity indices are driven by systemic, correlated crash risk that diversification does not neutralize. The mechanism is the opposite of idiosyncratic: macro shocks, liquidity crises, and correlated forced selling create co-crashes across all large-cap constituents simultaneously, preserving the heavy-tail behavior at the portfolio level. The same Gabaix-group’s earlier work on 1-minute S&P 500 returns reports α ≈ 2.75 for the negative tail of the index itself — identical to Taleb et al.’s 2.75 calibration value derived from the index options surface. The aggregation concern is valid in theory; the data do not support it in practice for equity indices.

The Structural Reason This Gap Persists

Options dealers cannot price at the correct power-law α without accepting hedging risk their balance sheets cannot carry.

Under a Pareto distribution with tail index α, the k-th moment exists if and only if k < α. For α < 4, the fourth moment of returns is infinite. The variance of a delta-hedging tracking error is proportional to E[(ΔS)⁴ · Δt²] — the fourth moment of the return increment scaled by time. When this fourth moment is infinite, the hedging error has no bounded expected cost per unit time; the standard Black-Scholes delta-hedging guarantee breaks down. The physical return α ≈ 2.75–3.0 for equity indices places them squarely in this zone. A dealer who priced deep OTM puts at the correct power-law level and hedged using power-law sensitivities would face unbounded tracking error; the standard dynamic replication argument fails exactly in the extreme-strike region where the mispricing is largest.

The rational response is to price at the hedgeable stochastic vol model, accept the resulting relative underpricing at extreme strikes, and collect the liquidity premium for providing markets in illiquid instruments.

The BIS Fundamental Review of the Trading Book (FRTB), “Minimum Capital Requirements for Market Risk,” January 2019 (BCBS d457) reinforces this incentive at the regulatory level. Under the Internal Models Approach (IMA), trading desks calculate market risk capital as Stressed Expected Shortfall at the 97.5% confidence level over a 250-day historical stressed window. A desk that prices deep OTM options with a correct power-law model — implying higher option sensitivities (Greeks) at extreme strikes — shows higher ES and therefore higher capital charges than a competing desk using stochastic vol with lighter implied tails at the same strikes. The specific mechanism: capital is calculated from sensitivity-weighted historical scenarios; higher Greeks at extreme strikes produce proportionally larger capital numbers under the same historical moves. The regulatory framework creates a systematic competitive incentive toward the thinner-tail calibration, making the gap self-reinforcing rather than self-correcting.

Capacity Analysis

The most defensible Universa figure is not the March 2020 headline number but the long-run portfolio result: a Wall Street Journal 2018 report found that a 3.3% Universa / 96.7% S&P 500 portfolio produced a 12.3% compound annual return in the 10 years through February 2018, compared to the index alone. That figure reflects real compound returns on a defined portfolio construction across a full decade including 2008 and 2011, and it is the number that conveys the strategy’s practical value for an allocator.

The March 2020 figures — a 3,612% return in March and 4,144% year-to-date per investor letters as reported by Bloomberg (April 8, 2020) — are expressed on required invested capital (the options premiums deployed as a fraction of the covered portfolio), not on total AUM. This is a non-standard denominator that amplifies percentage returns relative to conventional fund reporting; presented without that context, it distracts more than it informs. Both figures come from investor communications and are unaudited; they reflect the fund’s own performance attribution.

Per a finews.com April 2026 interview with COO Brandon Yarckin, citing the firm’s Form ADV filed with the SEC, Universa manages approximately $20 billion in Regulatory Assets Under Management since its 2007 founding. The primary Form ADV is public at adviserinfo.sec.gov (CRD 146052); RAUM for an options-focused manager may include covered portfolio notional rather than solely deployed premium.

The capacity ceiling — approximately $15–20B in covered portfolio notional per fund before market impact at extreme strikes materially closes the spread — is an analytical inference from the observable structure of SPX options markets: open interest at delta-5 and below runs approximately one to two orders of magnitude thinner than at delta-25, constraining the size of unidirectional positions before self-impact becomes the binding constraint. This estimate has not been formally quantified in any paper I am aware of; it should be treated as an order-of-magnitude inference, not a calculated bound.

Counterargument: This Is Just the Variance Risk Premium at a Different Strike

The VRP literature — Bollerslev, Tauchen, and Zhou (2009, Review of Financial Studies), Bondarenko (2014, Quarterly Journal of Finance) — documents that implied variance systematically exceeds realized variance, producing positive expected returns from variance-selling strategies. The objection is that the deep OTM relative underpricing is a manifestation of the general VRP — already widely known, already traded.

The distinction is structural, not semantic. The VRP is defined as the difference between risk-neutral and physical expectations of integrated variance — a scalar quantity computed across the entire distribution. It is earned primarily at ATM and near-OTM strikes in liquid instruments (variance swaps, short straddles, VIX futures). The deep OTM relative mispricing is a shape property of the vol surface: the question is whether the ratio of a delta-2 put price to a delta-15 put price is consistent with the power-law continuation from the near-money anchor. A vol surface can simultaneously have high integrated implied variance (high VRP) and an incorrectly extrapolated extreme tail — these are orthogonal properties.

The market structure confirms the distinction. VRP strategies operate in instruments of genuine liquidity — SPX variance swaps, short straddles, and VIX futures attract hundreds of billions in competing capital, compressing the premium continuously. Deep OTM puts at delta-2 to delta-5 are traded in markets with bid-ask spreads that can be several times wider than near-money options, and with open interest thin enough that large unidirectional positions face meaningful self-impact. The friction conditions that prevent full arbitrage of the deep OTM gap are precisely absent in the liquid near-money VRP trade. If these were the same edge, they would face the same competition and converge to the same premium. They don’t.

What Would Change This View

Two developments would close the deep OTM relative underpricing.

First: if options dealers adopted power-law tail pricing beyond the Karamata constant — calibrating extreme-strike options using the Hill estimator from physical return data rather than extrapolating stochastic vol models — the relative mispricing closes without requiring arbitrageur activity. The current barrier is hedgeability: dynamic replication fails when the fourth moment is infinite. A viable instrument for hedging tail-index risk itself, or a regulatory accommodation for bounded model risk in the extreme-strike book, would lift this constraint. Neither exists as of mid-2026.

Second: if sufficient competing capital entered the deep OTM long-put trade to overwhelm the thinness of the extreme-strike market — on the order of $50–100B in covered notional competing simultaneously — sustained buying pressure would push the market-implied α toward the physical estimate. The current scarcity of scaled practitioners is the structural condition that keeps the gap open.

The Actionable Implication

The monitoring signal is computable from public data. Estimate the physical α from the Hill estimator on the trailing five-year daily S&P 500 return series with k selected via bootstrap MSE minimization. Then compute the power-law continuation price for delta-5 and delta-2 puts, anchored to the observable delta-15 market price:

P(K_extreme) = P(K_anchor) × [(K_extreme − S₀) / (K_anchor − S₀)]^(1−α)

When market prices for extreme-strike puts fall materially below this power-law price — meaning the vol surface extrapolation uses an effectively higher α than the physical estimate — the relative underpricing is widest. The trade is long far-OTM S&P 500 puts at 3–6 month maturities, sized as a small fraction of the covered portfolio, held to expiry or monetized into sharp implied vol spikes.

The edge from the Hill estimator has migrated from the equity cross-section — where no structural barrier existed and post-publication decay was total — to the options market, where the hedgeability constraint means that even a well-resourced competitor cannot price away the wedge without accepting model risk their balance sheet cannot carry. The barrier that killed the equity cross-section alpha is precisely what is absent in the options-market application. That asymmetry is not coincidence — it is the structure of where durable edges live.

📊 Want Deeper Quantitative Analysis?

This research took very long time of data collection, verification, and analysis. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

Primary sources: Kelly-Jiang (2014, RFS) · Hou-Xue-Zhang (2020, RFS) · McLean-Pontiff (2016, JoF) · Gabaix et al. (2003, Physica A) · Taleb et al. (2019/2023, arXiv:1908.02347) · Bondarenko (2014, QJF) · Bollerslev-Tauchen-Zhou (2009, RFS) · BIS FRTB d457 (2019) · Chen-Lopez-Lira-Zimmermann (2022, arXiv:2212.10317)

Follow the research: YouTube — The Mathematical Trader · LinkedIn — Navnoor Bawa

Equity index option alpha, the compensation investors earned for selling volatility beyond what their market exposure alone would explain, has converged to statistically indistinguishable from zero over the past fifteen years, as the frictions that once kept non-dealer investors from selling options eroded and the burden of warehousing that risk spread beyond a small set of capital-constrained dealers. Crude oil’s variance risk premium has not been shown to follow the same path, but that claim needs to be stated precisely: the literature on it runs from 2013 through 2025, yet no published study has re-tested whether the premium’s defining number, a roughly 40% Sharpe ratio for a diversified short-variance portfolio, still holds outside the 1989–2011 sample it comes from. Absence of a decay study is a gap in the evidence, not proof the edge survived intact, and the rest of this piece tries not to blur that distinction. What the evidence does support is narrower and still useful: capturing this premium requires a full options surface and balance-sheet access that remain genuinely scarce, a structural barrier rather than an informational one, and that barrier is also the reason no one has published a capacity estimate for the trade. This article sets out that mechanism, what the literature does and does not say about decay, and what can and cannot be inferred about capacity from the size of the market the trade would have to clear through.

🎬 Prefer to watch rather than read? A NotebookLM-generated video overview of this article is available here: Watch the video overview → Full analysis, citations, and data remain in the article below.

The consensus, stated precisely

Using options on 21 commodities from 1989 to 2011, Prokopczuk and Wese Simen construct synthetic variance swaps and document significantly negative variance risk premia in 17 of 21 markets at 60 days, meaning implied variance systematically overpays for the realized variance that follows. A diversified short-variance portfolio earns an annualized Sharpe ratio near 40%, roughly four times a passive long-futures portfolio over the same sample; energy-sector Sharpe ratios at the two-month horizon range from 34% to 47%. The result survives interpolation-method changes, alternative truncation points, jump-risk adjustment, transaction costs up to 5% of the swap rate, and a 20% performance fee. This is a well-established empirical fact in the derivatives literature. The questions an institutional reader actually needs answered are why it has not been arbitraged away, and how much capital could chase it before it was.

What killed the equity index version, and why it hasn’t happened here

The clearest available decay study is not about commodities. Dew-Becker and Giglio, in a 2025 Chicago Fed working paper, document that traded equity index option alphas have converged to statistically indistinguishable from zero over the past 15 years; their general statistical tests place the break around 2010, and a more targeted test using net dealer gamma exposure dates the specific shift to 2012m5. Their explanation is not simply that dealer balance sheets grew; it is that the frictions preventing non-dealer investors from supplying (selling) options declined, so the asymmetric risk-bearing burden that used to fall on dealers spread across a wider pool of capital, including the growing hedge fund sector and structured-product issuance to retail. The paper documents this directly: net dealer gamma exposure in S&P 500 options went from consistently negative before 2012 to roughly zero or positive after, with timing the paper describes as highly similar to the timing of the shift in the option premium. This is the textbook decay path for a known anomaly: a structural inefficiency exists, the frictions that prevented competition from closing it erode, and the alpha goes to zero.

No equivalent decay study exists for crude oil, and that needs separating from a different, true fact: the literature has kept extending, just not on the question of whether the original 40% Sharpe ratio replicates in fresher data. Li and Li (2025) decompose crude oil’s tail risk premium from its variance risk premium and find the tail component, though smaller in magnitude, carries more significant predictive power for futures returns. Ammann, Moerke, Prokopczuk and Würsig (2023) go further: across commodity markets generally, both left- and right-tail risk implied by options are economically large, a contrast with the equity-options literature, where the Dew-Becker and Giglio data above shows risk aversion rising sharply on the downside specifically, the familiar left-skew, crash-premium pattern. The Ammann et al. paper adds a second, separate finding directly relevant here: left and right tail risk are largely independent of the variance risk premium itself, the academic confirmation of the mechanical point above that tail/skew risk and the level of implied variance are different objects, priced separately, and a single 30-day point estimate cannot carry information about one while measuring the other. None of this is a test of whether the 1989–2011 Sharpe ratio survives on, say, 2015–2025 data. That specific test does not appear to have been published. Its absence should be read as an open question the data-access barrier may itself help explain, since replicating it requires the same options-chain data that limits practitioners, not as evidence either way about whether the edge has decayed.

The mechanistic reason the two markets diverged is access, not attention. Closing the equity VRP required frictions on non-dealer option supply to erode, a process that unfolded steadily as structured-product access for retail and the hedge fund sector both grew after the financial crisis. Closing the commodity VRP requires something narrower: a full strike-and-maturity options surface, sourced from a paid vendor, to replicate the variance swap and read the skew that prices the tail. CBOE’s OVX collapses that entire surface into one 30-day number computed from near-month USO options. CME’s CVOL index computes the actual skew and convexity from the full curve, proving the surface problem is solvable, but distributes the result through several licensed channels, live streaming via the CME Market Data Platform and CME Direct, historical data via CME DataMine, and programmatic access via a paid REST API, every one of them gated behind a paid account entitlement. A capital constraint eases as more capital arrives. A data-access constraint does not ease just because more people want the data; it eases only when someone pays for it. That is a structurally different, and more durable, kind of barrier, and it is the direct explanation for why the commodity premium has not traced the equity premium’s path to zero.

What would actually constrain capacity, and what doesn’t

The absence of a published capacity estimate for crude oil variance harvesting is itself informative: capacity research gets written about trades that are crowded enough to need it. Value, momentum, and carry in commodities have exactly that literature. Kang, Rouwenhorst and Tang construct a crowding measure directly from CFTC positioning data and show it has a strong negative predictive impact on those factors’ expected returns, with historical factor returns accumulated primarily during periods of low crowding. No comparable study exists for commodity variance harvesting, because the data barrier above has kept the population of participants who can correctly measure and size the trade small enough that crowding has not yet become the binding constraint research needs to explain.

That does not mean the ceiling is infinite, and the outer bound can be reasoned from public market-size data even without a published capacity study. CME states that NYMEX WTI futures and options trade over 1 million contracts daily, against roughly 4 million contracts of open interest. CFTC’s own reporting shows two different numbers depending on which report is pulled, and the gap between them is itself instructive. The Disaggregated Petroleum Combined report, which nets futures and options together and is the more relevant measure for a variance-replication strategy, shows the WTI-Physical contract at 3,032,488 contracts of open interest as of the May 12, 2026 reading; as of this writing (June 20), that combined report has not updated past May 12 despite CFTC’s stated weekly cadence. The separate futures-only report for the same contract has continued updating normally, showing 2,025,180 contracts as of June 2, 2026. The two figures are not directly comparable, one nets in options exposure and one does not, but the combined report’s monthlong gap is a live, current example of the exact problem described later in this piece: the data series that actually matters for sizing options-based risk lags worse in practice than its official cadence promises. At a WTI price near $78 per barrel as of June 19, 2026 (CNBC), the combined figure alone represents approximately $237 billion in notional exposure. That headline figure is the outer bound, not the relevant one: it is dominated by liquid near-the-money strikes, while variance-swap replication needs depth specifically in the out-of-the-money wings, and the share of total open interest concentrated there is not published in any aggregate public series (magnitude not obtainable from public data).

A widely cited example of how short-volatility capacity constraints can fail is the XIV and SVXY collapse, but the mechanism behind it needs to be stated precisely before it can be borrowed for a different market. By early 2018, the two largest short-volatility exchange-traded products, XIV and SVXY, held a combined notional position of roughly $280 million in VIX futures vega, against a total VIX futures market notional of roughly $7 billion the prior year. Augustin, Cheng, and Van den Bergen, in the peer-reviewed account published in the Financial Analysts Journal, document how the ETPs’ daily rebalancing mandates forced them to buy VIX futures into the closing minutes precisely as volatility was spiking, mechanically increasing the size of the next required purchase in a self-reinforcing feedback loop. Independent tracking of the session by volatility analyst Vance Harwood found that a combined $4 billion of the two front-month VIX futures contracts changed hands at the close, as the VIX itself closed up roughly 116% on the day. XIV lost over 90% of its value within hours; Credit Suisse announced its termination the next day, with trading continuing until February 15 and final cash settlement on February 21, sixteen days after the crash.

The forced-selling mechanism here was specific to the product structure, not to being short volatility in general. XIV and SVXY were exchange-traded notes with a contractual, daily, price-insensitive rebalancing mandate: by prospectus, they had to trade toward a constant leverage target every day regardless of where the market was, which is what created the reflexive loop. A hedge fund running discretionary or systematic variance-swap replication has no such covenant. It can widen hedging bands, cut size, or stop trading into a dislocation in a way an ETN legally cannot. Importing the “fails discontinuously, not gradually” conclusion from Volmageddon without checking for that precondition would be reasoning by analogy dressed up as derivation, and it is worth naming the better-grounded mechanism instead.

That mechanism is margin and VaR-driven deleveraging, documented well outside the ETP world. Brunnermeier and Pedersen model what they call the margin spiral and the loss spiral: as volatility rises, a position’s value-at-risk rises with it, prime brokers raise margin requirements, and a fund is forced to cut the position to stay within risk limits, which itself pushes prices further and tightens margins again. This is a real constraint on actual funds, not a contractual artifact of an ETP structure, and it is the closest thing to a generalizable mechanism for how a short-oil-variance book could be forced into procyclical selling exactly when liquidity is thinnest. But naming the mechanism is not the same as showing it binds here: no published source documents a margin or VaR-driven unwind specific to an oil variance-replication strategy, at any size, and the capital and risk-limit terms that would actually trigger one are fund-specific, private, and not observable from outside.

That gap connects to a separate, plainer problem: no public track record of a fund actually running this trade exists. Every figure in this piece comes from an academic dataset or a dead VIX product, not from a disclosed P&L. That absence cuts both ways. It is consistent with the data-access barrier described above, if the trade were being run successfully at scale by identifiable participants, some trace of it would likely be visible by now, in marketing materials, a research note, or a recognizable positioning pattern, and none is. But it also means every claim in this section about capacity is reasoning from market size and an adjacent but structurally different blowup, not from observed behavior of the actual trade. The honest statement is that a real mechanism for capacity constraints exists in principle, margin- and VaR-driven, not contractually forced the way XIV’s was, and that the data needed to size it for crude oil variance harvesting specifically, OTM wing depth, fund-level risk limits, or a disclosed track record, is not publicly available. That is a narrower claim than a derived ceiling, and it is the one the evidence supports.

COT cannot substitute for the crowding signal it could in principle provide

The Kang, Rouwenhorst, and Tang result above establishes something specific: CFTC positioning data, used correctly, does carry real predictive power over forward commodity factor returns. That makes its limitations for this purpose a sharper problem than a generic data-quality complaint. The Commitment of Traders report is published every Friday for positions held the preceding Tuesday, a built-in three-day lag, and by law the CFTC does not disclose how individual traders are classified within each reporting category. In practice the lag is sometimes worse than the design: the combined futures-and-options report cited above was, as of this writing, still anchored to a May 12 reading more than five weeks stale, even as the futures-only report for the same contract kept updating weekly. A crowding measure that is structurally lagged, intermittently more so than advertised, and category-level rather than participant-level can flag a slow build in aggregate positioning. It cannot tell a risk manager, in real time, whether their own variance-replication flow is approaching the kind of liquidity fraction that broke the VIX futures market in 2018. The data exists in principle to monitor this capacity ceiling. The free version of it is built for weekly macro context, not real-time risk management, and even that weekly cadence is not always honored for the specific report type that matters most here.

The 2026 Hormuz episode, including the de-escalation

Brent opened January 2026 near $60.75, reached $116.29 on March 9 as fighting closed the Strait of Hormuz to tanker traffic, fell to $80.33 the following day, and recovered to roughly $105.38 by mid-May (Capital.com). That alone is the whipsaw a short-variance position is exposed to. But the episode kept moving after mid-May. Trump and Iranian President Pezeshkian signed a memorandum of understanding at the Palace of Versailles on June 17, halting hostilities for a 60-day negotiating window, and crude fell nearly 25% over the following week as the geopolitical premium came out of the market (Capital.com). Two days later, the first round of follow-up technical talks, scheduled for June 19 at Switzerland’s Bürgenstock resort and meant to begin converting the interim truce into a lasting settlement, were postponed indefinitely, with the Swiss Foreign Ministry citing unresolved logistics rather than a breakdown in the agreement itself. OVX, which peaked at 125.99 during the crisis, closed at 51.54 on June 18 (Yahoo Finance), reflecting the de-escalation but still well above its pre-crisis 52-week low of 23.59.

This second move is, if anything, a sharper illustration of the argument than the original spike. A trailing single-point OVX read taken any day in this sequence, the March peak, the March crash, the May recovery, or the post-MOU plunge, would have told a researcher the level of 30-day implied volatility and nothing about which of those regimes they were actually in or how durable it was. The Bürgenstock postponement makes that concrete within 48 hours of the MOU itself: a peace agreement and a stall in the talks meant to cement it landed within the same week, and a position sized off any single day’s OVX print in between would have had no way to distinguish a market pricing durable peace from one pricing a truce that could still unravel. The interim MOU is real and the price move it caused is real, but the postponed talks mean the underlying political risk has not resolved, only repriced, and a single 30-day number cannot tell a risk manager whether the next move is further decay toward pre-crisis levels or a snap back if the negotiating window stalls. CBOE’s own volatility research desk documented the same mechanism during a comparable, smaller 2025 Iran-linked spike: WTI one-month implied volatility jumped to 68% before easing to 51% within the same week, while the implied-realized spread, the premium a short-variance position is actually paid, compressed from 30 points to 14 (CBOE). The premium compresses fastest exactly when a position needs it most, in both directions, and that is true whether the regime shift is an escalation or a de-escalation.

The mechanism note that builds the variance-swap replication formula from first principles — including the vol-squared derivation behind the 59%/83% gap above and the three data acquisitions needed to size against it properly — is on Patreon.

The obvious objection

OVX has documented predictive power for subsequent realized oil volatility, in-sample and out-of-sample, raising a fair question: why isn’t that good enough to time entries into the premium? Forecasting the average level of realized variance and constructing a hedgeable short-variance position are different tasks. Forecasting skill says implied volatility is rich or cheap on average; it says nothing about the price of the convexity being sold, which is where the tail-risk literature above places the loss exposure, and which only shows up in strike-level surface data. The underlying instrument compounds this: USO near-month options, what OVX is built from, thin out precisely in the deep-wing strikes that matter most during a stress event, a liquidity constraint separate from data availability.

What would change this view

Two specific findings would falsify the claims made here. First, a published study showing crude oil variance-swap-replication alphas converging toward zero over a recent sample, the commodity equivalent of the Dew-Becker and Giglio equity result, would mean the data-access barrier has closed faster than the literature reviewed here suggests. No such study currently exists. Second, a published capacity or crowding estimate specific to commodity variance or volatility-selling strategies, comparable to the Kang-Rouwenhorst-Tang result for value, momentum, and carry, would mean the trade has become crowded enough to study, which would also mean it is closer to its ceiling than the absence of such research currently implies. Both are reasonable things to expect eventually; neither has been published yet, and that gap is the actual state of the evidence rather than a claim about what it will always show.

The actionable implication

A fund without a full options surface, faster positioning data, and point-in-time fundamentals should treat its oil volatility research as regime-level and directional, not as a sized short-variance strategy with a stated Sharpe ratio. That data barrier is real and is the most defensible reason the edge has persisted in the published record. For a fund that does have the full surface and balance-sheet access to implement variance-swap replication properly, the binding constraint shifts from data to liquidity and risk-capital terms: position sizing needs to be set against the depth of the specific OTM wing strikes being traded, not against the headline open-interest figure, and against margin and VaR terms that tighten exactly when realized volatility spikes, particularly through a regime shift of the kind the 2026 Hormuz episode produced twice in four months, first the escalation, then the de-escalation, where wing liquidity disappears fastest in both directions. No public source states what fraction of available liquidity, or what margin terms, would actually bind for this trade. That is a real gap in what can be said here, not a number this piece is able to round to. What the public record does support is narrower: margin- and VaR-driven deleveraging is a documented feature of how short-volatility books fail under stress, and whoever is capturing this premium today is doing so without a published capacity study to size against.

Further reading: The full mechanism note covers the static replication formula (Britten-Jones-Neuberger), the CFTC combined-report staleness problem, the margin-spiral capacity mechanism (Brunnermeier-Pedersen), and the three specific data acquisitions that close the gap from the 1989–2011 academic finding to a sized position in 2026.

📊 Want Deeper Quantitative Analysis?

This research involved extensive data collection, source verification, and multi-round fact-checking against primary sources. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon — the same rigour as this piece, applied to ideas that don’t make it into public articles.

By joining, you will be supporting independent research and motivating me to publish more content like this.

→ Join the Patreon community here

For more quantitative finance research, subscribe to The Mathematical Trader on YouTube · Connect on LinkedIn

🎬 Prefer to watch rather than read? A NotebookLM-generated video overview of this article is available here: Watch the video overview → Full analysis, citations, and data remain in the article below.

The Consensus and Its Missing Variable

The dominant published view on PE in European football is accurate as far as it goes: clubs are globally recognized brands trading at 3–6x revenue against peer entertainment assets at 10–12x, implying a structural discount. According to PitchBook’s Big Five football dashboard, more than 36% of Big Five clubs have financial backing specifically from PE, VC, or private debt firms at the start of 2025-26; multi-club ownership networks (MCOs) — a broader category that includes strategic investors and family offices alongside PE — cover 48% of Big Five clubs, up from 41.7% in 2024. Source limitation: PitchBook’s figures reflect publicly disclosed deal information; undisclosed private stakes are not captured and the true penetration rate is likely higher.

European football dealmaking grew from approximately €66.7 million in 2018 to a peak of approximately €4.9 billion ($5.4 billion) in 2022, the year the Chelsea and AC Milan transactions settled, per the PitchBook Q3 2023 Analyst Note. Deal activity has remained elevated since but has not exceeded the 2022 peak, according to PitchBook’s Part III update from August 2025.

What the consensus omits is the distribution within the sector. A club with €600M in revenue priced at 5x EV/Revenue and a club with €80M in revenue priced at 1.3x are not the same trade. Treating them as one asset class generates the illusion that the opportunity is either uniformly attractive or uniformly crowded. It is neither.

The IRR Math: Why Trophy and Mid-Tier Are Different Trades

All documented inputs are cited to primary sources; exit multiples are scenario assumptions explicitly labeled as such.

Documented inputs:

• Revenue CAGR: 6.2%, for Europe’s top 32 clubs since Football Benchmark’s inaugural 2016 edition through 2025. Limitation: Football Benchmark is a commercial analytics firm; its figures derive from club-reported financials with variable auditing standards across leagues.

• Five-year hold (standard PE football horizon).

Scenario A — Mid-Tier Entry at 1.5x Revenue:

Exit multiple: 3.5x — an assumption reflecting FSR-forced EBITDA normalization and unlocked UCL access. There are no documented PE football exits yet to anchor this figure; it is a scenario input, not an empirical estimate. The first exits from the 2018-2021 cohort will validate or substantially revise it. If only FSR normalizes and UCL access is not achieved, a 2.5x exit yields approximately 17.6% gross IRR. The model is sensitive to this assumption in both directions.

Scenario B — Trophy Entry at 5x Revenue:

Exit multiple: 6x — an assumption of modest expansion from current levels. This is below Manchester United’s documented 6.5x in Football Benchmark 2024 and therefore conservative for the trophy tier, though the 6x exit is itself undocumented for any completed PE football transaction.

Net IRR to LPs after fees: approximately 6–7%. This does not clear the institutional PE hurdle rate for a $5B+ fund. The 6x exit assumption is generous given Accuracy Group’s analysis of Chelsea (2022) and Manchester United INEOS (2024) shows actual transaction prices ran approximately 20–30% above Football Benchmark’s theoretical valuation — buyers at trophy level paid a control premium before the hold period began.

Scenario C — Mid-Tier Entry → Relegation:

Serie A domestic rights for 2024-2029 total €900M/year across 20 clubs, per Reuters’ reporting of the official October 2023 club vote, giving a mid-table club approximately €40–55M annually. Serie B’s total broadcast rights for the 2021-2024 cycle stood at €48.5M for the entire league, per Sportcal’s direct reporting of the Sky + DAZN deal. Limitation: this figure covers the 2021-24 cycle; the current cycle may differ. A primary source for current per-club Serie B distributions was not obtainable. The structural argument — that any individual Serie A club’s annual broadcast share substantially exceeds a pro-rata share of a ~€50M total pool — holds regardless of the exact current figure.

Post-relegation revenue falls toward approximately €60M (broadcast line collapses; other revenues partially hold). Distressed exit at 1x revenue implies approximately €60M EV.

Equal capital split between Scenario A and Scenario C produces a blended MOIC of approximately 1.78x (€532.9M returned against €300M invested) — a materially positive outcome, not the netting-to-zero that a naive average of the two gross IRRs (25.8% and -26.3%) might suggest. That naive average is the wrong calculation: it treats percentage returns as additive across mismatched hold periods (5 years versus 3) and ignores capital weighting. The portfolio implication is the opposite of intuitive — the trade survives even a 50% Scenario C hit rate. What it does not survive is Scenario C clustering non-randomly across a portfolio rather than landing independently, which is the more realistic underwriting concern, and the subject of the next section.

A Case Consistent With the Thesis: Hellas Verona

In January 2025, Presidio Investors acquired Hellas Verona, as confirmed by the joint statement published by the club and Presidio. Financial terms were not officially disclosed. Private Equity Wire, citing a source familiar with the deal, estimated the enterprise value at €120M–130M including existing debt. TheScore.com citing AFP reported Italian media estimated approximately €130M total. According to the managing partner in an interview with ION Analytics/Mergermarket — a practitioner’s statement, not independently verified — no new debt was raised; Presidio bought out the existing equity with existing debt remaining in place.

Hellas Verona was officially relegated from Serie A during the 2025-26 season, per Football Italia’s final standings report, confirmed independently by Yahoo Sports. The club finished 19th — within 17 months of acquisition. The managing partner stated, per the ION Analytics interview, that the firm reviewed approximately “70 to 80 different clubs” before selecting Verona.

This is a single case (n=1). It is consistent with the thesis that relegation risk is systematically underweighted in PE underwriting of mid-tier football clubs. It is also consistent with Presidio simply being unlucky, or with them being poor at evaluating squad quality for survival. One observation cannot establish a pattern. What it does demonstrate is that the tail event is accessible even to investors who conduct thorough deal selection; the number of clubs reviewed is not the relevant variable. The academic evidence below argues the mechanism is structural, not random — but that argument rests on limited evidence and should be weighted accordingly.

Why the Mid-Tier Discount Exists and Why It Is Changing

Three structural factors price mid-tier clubs below their asset value: illiquidity (no exchange-traded market), relegation risk (can destroy 40–60% of EV within a season), and governance weakness (family/founder ownership with no institutional cost discipline). These are real and explain the discount.

Two catalysts are changing the underlying economics in ways backward-looking multiples do not capture.

The FSR Inflection: UEFA’s regulations cap squad costs at 90% in 2023/24, 80% in 2024/25, and establish a permanent 70% ceiling from 2025/26. Football Benchmark 2025 documents the average squad cost-to-revenue ratio declining from 95% in 2023 to 82% in 2025. A club with €100M revenue and squad costs structurally capped at €70M posts a visible forward EBITDA margin for the first time. Standard DCF frameworks will update when 12–18 months of constrained financial statements are published.

The UCL Media Rights Cycle — What the Actual Bids Show: The 2027–2033 tender, managed by Relevent Sports under a mandate announced in March 2025, has produced disclosed results for the big five markets. Bloomberg reported in November 2025 that UEFA will receive approximately €2.5 billion per year from the five major European markets in the next cycle — up from approximately €2 billion in the current cycle, a 25% increase. The total revenue target across all competitions, all territories, and all revenue streams (broadcast, sponsorship, and licensing) exceeds €5 billion per season, per the UC3 joint venture’s stated goal — roughly 14% above current total men’s club competition revenue of €4.4 billion, per UEFA’s own 2024-25 financial results.

The 25% increase in big five media rights is meaningful and confirmed. At a constant 4.9x EV/Revenue multiple, a €20M annual UCL distribution increment for a mid-tier club that achieves qualification translates to approximately €100M in EV — significant option value for a club purchased at 1.5x revenue, even without the secondary catalysts.

The Mechanism That Makes Scenario C Non-Random

The most analytically interesting feature of the Hellas Verona case is not the outcome but the path: Presidio’s value creation strategy would have required rotating to younger, cheaper players to achieve FSR 70% compliance, which simultaneously disrupts the player coordination structures that determine on-field performance. If this observation is general — if the FSR compliance mechanism that drives the financial upside also degrades the squad quality that prevents relegation — then Scenario C is not an independent tail event. It is endogenous to the strategy.

This is a hypothesis, not a demonstrated mechanism. The evidence supporting it is preliminary: a 2024 SSRN working paper by Kristina Lalova-Yuhasz (Michigan State University) studies 96 European clubs (28 PE-backed, 68 non-PE) across the five major leagues in a staggered difference-in-differences design and finds PE investments increase commercial and matchday revenue but decrease player network centrality metrics and on-field match performance. Disclosure: the SSRN submission lists Lalova-Yuhasz as sole contact author; ResearchGate’s indexing of the paper lists Brennan Cimpeanu as co-author alongside Lalova — the same attribution Cimpeanu makes on his LinkedIn profile. The conflict between SSRN (sole author) and ResearchGate (two authors) is not resolvable from public information. ECGI’s publication of the author’s summary captures the tension: PE is “highly effective in enhancing financial outcomes” while this “comes at the expense of on-field performance.” Note: this is Lalova-Yuhasz’s own ECGI blog post, not an independent editorial assessment.

The limitations are significant: pre-peer-review, single season (2023-24), 28 PE-backed clubs, mechanism not externally replicated. The endogeneity claim does not require the Lalova-Yuhasz paper to be correct — it is a structural observation about the strategy’s internal logic. The paper provides preliminary empirical support, not proof. If the finding survives peer review and replication across multiple seasons, the probability that Scenario C is endogenous rather than random increases substantially. If it does not survive, the mechanism reverts to a structural hypothesis with one consistent anecdote.

The practical implication is the same either way: a PE investor who deploys the standard “optimize costs, rotate squad, improve commercial” playbook without modeling the joint probability distribution of FSR compliance and relegation avoidance is misspecifying the return distribution.

The Crowding Signal: Where Alpha Has Already Decayed

The trophy-asset tier is clearly crowded. Accuracy Group’s analysis of Chelsea (2022) and Manchester United INEOS (2024) shows actual transaction prices approximately 20–30% above Football Benchmark’s theoretical valuation. Paying a control premium on an EBITDA-negative asset at 5x revenue is the signature of more capital than deals.

The crowding is accelerating. Named vehicles with documented European football investments or public commitments now include: Apollo Global Management ($5B Apollo Sports Capital, September 2025), Ares Management (sports fund launched 2022, reportedly building a second), CVC Capital Partners (refinancing a sports portfolio valued at £9 billion as of July 2025 per Bloomberg and Sportcal — a PitchBook article from November 2025 cited “$14 billion” for the same vehicle; the discrepancy may reflect portfolio revaluation or different asset inclusions and is not resolvable from public information), RedBird Capital Partners, Clearlake Capital, Oaktree Capital Management, Sixth Street Partners, Arctos Partners, MSP Sports Capital, TPG, Presidio Investors, and Elliott Management. This is approximately 12 vehicles based on publicly confirmed transactions and announcements, and is a lower bound — undisclosed vehicles are not captured. The August 2024 NFL PE approval additionally admitted Blackstone, Carlyle, Dynasty Equity, and Ludis to American sports, expanding the pool of institutional capital with sports-sector exposure.

The Regulatory Barriers That Permanently Reduce the Investable Universe

Germany — 50+1 Rule: In July 2025, the German Federal Cartel Office issued preliminary findings that the Bundesliga’s 50+1 ownership rule is broadly compatible with competition law, while requiring the DFL to enforce open club membership more strictly. The FCO’s guidance is preliminary, not a final ruling, but the practical effect is unchanged: PE cannot hold control positions in Bundesliga clubs, removing approximately 18 Bundesliga 1 teams from the mid-tier PE control thesis.

England — Football Governance Act 2025: The Act, receiving Royal Assent in July 2025, established the Independent Football Regulator with powers to approve ownership suitability, impose licensing conditions on financial planning and fan engagement, and intervene in revenue distribution. Licensing implementation phases throughout 2026, extending deal timelines and constraining operational autonomy for Premier League clubs.

France — Ligue 1 Structural Revenue Crisis: Deloitte’s Football Money League 2026 confirms the DAZN termination “will negatively impact French clubs’ broadcast revenues in the short-to-medium term.”

Capacity Bound: A Transparent Construction

A precise count of qualifying mid-tier clubs is not publicly available. What follows is a transparent derivation from documented inputs, not a database count, and should be read as an order-of-magnitude estimate.

Starting universe: Five Big Five leagues, approximately 96 clubs total.

Filter 1 — Regulatory inaccessibility: Germany’s 50+1 rule removes approximately 18 Bundesliga 1 clubs from PE control consideration. France’s broadcast crisis excludes Ligue 1’s 18 clubs from the near-term accessible universe — not a permanent barrier like Germany’s, but treated the same way for this construction. The accessible universe for control-position mid-tier investments concentrates in England, Italy, and Spain — approximately 60 clubs (96 total, minus 18 Bundesliga and 18 Ligue 1).

Filter 2 — Already PE-backed: PitchBook’s dashboard reports 36% of Big Five clubs have PE/VC/private debt backing. Scaling that penetration rate to the 60-club accessible universe (acknowledging this is an approximation, since the German penetration rate is near zero and distorts the Big Five average) suggests approximately 20–25 accessible clubs are already PE-backed, leaving approximately 35–40 uninvested clubs in accessible leagues.

Filter 3 — Trophy-tier pricing: The top 6–8 Premier League clubs trade at 4x+ revenue (trophy tier, Scenario B territory). Removing these leaves approximately 28–34 mid-tier clubs in accessible leagues currently available for entry at 1–2x.

This construction produces an estimate of approximately 30 qualifying clubs with a range of roughly 25–35. That range is wide enough that the capacity claim cannot be stated precisely. What can be stated: the universe is not large, and with approximately 12 named vehicles actively pursuing it, the deal-to-buyer ratio is thin.

Fund-size implication: A dedicated sports vehicle with $500M–$2B AUM can make 5–10 investments at €150M average check without being forced into trophy assets. At $5B+ AUM, a fund cannot deploy at scale into mid-tier clubs without concentration risk and is functionally pushed into Scenario B. The alpha at mid-tier is structurally inaccessible to large-cap generalist PE — not because of any analytical advantage the smaller funds have, but because they can match their fund size to the available deal flow.

→ A condensed, single-document version of this analysis — the same sourced figures and IRR math, formatted as an institutional reference note — is available on Patreon.

What Would Change This View

UCL big-five media rights disappointing further. The €2.5B/year from big five markets is confirmed for 2027-2031. If subsequent tender waves (Eastern Europe, Americas, Asia-Pacific) yield materially below expectations, total revenue fails to reach the >€5B target and the incremental club revenue uplift is smaller than the 25% already achieved in big-five media rights.

UEFA FSR enforcement asymmetric. The 70% permanent cap must produce documented sporting sanctions — not just fines — against non-compliant clubs in 2025-26 for the EBITDA normalization to materialize.

The endogeneity hypothesis failing to replicate. If peer-reviewed replication of the Lalova-Yuhasz finding shows no systematic on-field decline under PE, or if it shows the effect is fully absorbed by clubs without relegation consequences, Scenario C reverts to a random tail event, the IRR distribution improves, and the mid-tier trade becomes more attractive than Scenario A suggests.

The 2018-2021 exit cohort disappointing. The first realized IRRs from PE football exits will either validate Scenario A assumptions or indicate the exit multiple assumed is too high. If realized exits cluster below 2.5x MOIC, the mid-tier thesis is under-earning and will not attract subsequent institutional capital on the necessary terms.

Actionable Implication

The mid-tier European football club trade has a closing window, the length of which is genuinely uncertain. The capacity construction above suggests 25–35 qualifying clubs remain uninvested; against approximately 12 named vehicles, the supply-demand balance is thin but not exhausted. The trade is accessible to vehicles of $500M–$2B with football operational expertise. The two observable signals worth tracking: (1) UEFA FSR enforcement producing documented sporting sanctions against clubs exceeding the 70% squad cost cap — fines without sanctions mean the cap is not binding; (2) the outcome of post-relegation revenue recovery for Hellas Verona under Presidio’s ownership, which will be the clearest near-term single-club test of whether the operational thesis can survive a Scenario C event.

Deloitte’s Sports Investment Outlook identifies 2025 as the year of first significant PE exits. Those realized returns, as they become public through LP reporting, will be the first empirical test of whether the trade delivered what buyers claimed.

📊 Want Deeper Quantitative Analysis?

This research took considerable time to collect, verify, and analyze. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

→ Join the Patreon community here

Navnoor Bawa writes at navnoorbawa.substack.com · YouTube · LinkedIn.