The most celebrated equation in finance maps exactly where markets misprice risk. This article covers the academic proof of the Variance Risk Premium, the delta-hedging P&L mechanics every vol desk runs, and the documented trades — convertible bond arbitrage, capital structure arbitrage, tail-risk hedging, and dispersion — that five decades of the world’s best hedge fund returns are built on.
Black-Scholes: The Quoting Language Every Options Trader Uses — and the Edge Hidden in Its Five Broken Assumptions
Fischer Black and Myron Scholes built their 1973 model on five premises: constant volatility, continuous trading, no dividends, no transaction costs, and log-normally distributed returns. Every professional derivatives trader knows these premises are false. The alpha — persistent return above market benchmarks — comes entirely from one activity: measuring how false each premise is, in which instrument, at which moment, and positioning before the rest of the market does.
Haug and Taleb established this formally in a 2011 Journal of Economic Behavior & Organization paper: sophisticated options traders do not use Black-Scholes-Merton to determine whether an option is fairly priced or to hedge positions. They rely on heuristics — often pre-dating BS — to convert raw prices into implied volatility numbers that can be compared across strikes, maturities, and underlyings. The actual trade is always in the gap between the implied volatility these heuristics produce and the realized volatility a better model forecasts.
Haug and Taleb identify the precise operational failure of the model: dynamic hedging — the core replication argument underpinning Black-Scholes — is not feasible in real markets because jump risk dominates. A 25% overnight gap in Ericsson, one of the world’s most liquid stocks and documented in the paper, can wipe out hundreds of weeks of delta-hedge gains. That failure is not a theoretical curiosity. It is a recurring opportunity for funds positioned to benefit from it.
The Variance Risk Premium: Academic Proof That Implied Volatility Systematically Overstates Realized Volatility
Before examining specific funds, the mechanism needs its empirical foundation.
Gurdip Bakshi and Nikunj Kapadia’s 2003 paper in the Review of Financial Studies, “Delta-Hedged Gains and the Negative Market Volatility Risk Premium”, is that foundation. Using S&P 500 index options data spanning January 1988 to December 1995 — split into two subsamples for robustness — they measured the P&L of a rigorously executed position: buy an SPX option, then delta-hedge it daily against the underlying index. Strip out directional exposure entirely. What remains is a pure volatility bet.
The result was unambiguous: the delta-hedged long strategy consistently underperforms zero. Option buyers systematically overpaid for volatility relative to what subsequently materialized. Crucially, the underperformance is greater during periods of higher volatility — precisely when the cost of the protection appears most justified — and is statistically robust across maturities. The option seller, running the mirror position, consistently collected the difference between implied and realized volatility.
Peter Carr and Liuren Wu extended this in their 2009 Review of Financial Studies paper, “Variance Risk Premia”. Their key finding: the Variance Risk Premium (VRP) is not explained by standard CAPM betas. It is an independently priced risk factor — meaning it carries a premium that equity exposure alone cannot replicate. This is the theoretical justification for why volatility-selling generates returns that are structurally distinct from simply owning stocks.
The practical scope: implied volatility exceeds realized volatility approximately 85% of the time — an empirical industry rule of thumb documented across multiple market regimes. That persistent structural gap — with losses clustered precisely in the crashes that BS’s log-normal distribution least anticipates — is the risk-return profile every institutional options desk in the world is operating around.
The Delta-Hedging P&L Equation Every Volatility Desk Runs
A delta-hedged short option position’s daily P&L follows directly from the BS framework:
Daily P&L ≈ Theta·Δt + (1/2)·Gamma·(ΔS)²
For a short-gamma (option-seller) position:
Theta = positive daily income from time decay
Gamma term = negative (losses scale with the square of underlying moves)
Therefore:
P&L > 0 when realized moves are smaller than implied → theta wins
P&L < 0 when realized moves are large → gamma losses exceed thetaExpressed in core BS terms: the hedged P&L is proportional to (RV² − IV²) · Gamma · Δt. For a short-gamma seller, this is positive whenever realized variance is below implied variance — the condition the Bakshi-Kapadia evidence shows holds the majority of the time. The volatility arbitrage mechanism is: sell overpriced implied vol, delta-hedge to neutrality, collect the IV-RV spread as options decay. The friction reality matters too — bid-ask costs, financing, margin requirements — which is why this edge accrues primarily to large shops with prime brokerage infrastructure and not to retail vol sellers.
This is the generic engine. The three strategies that follow each exploit a specific dimension of BS’s failure: embedded option mispricing in convertibles, the credit-equity divergence that BS cannot see, and the fat tails that BS structurally underprices.
Ken Griffin and Citadel: Convertible Bond Arbitrage and a Pricing Model Now on Version 600
The clearest practitioner account of how the BS formula creates an exploitable map begins in the fall of 1987, in a Harvard dorm room.
In a Stanford GSB interview, Griffin described the origin: he needed real-time pricing to engage in arbitrage between common stocks and related derivatives — convertible bonds, warrants, preferred securities. This is why the 18-year-old Harvard sophomore arranged to have a satellite dish installed at Cabot House. As Institutional Investor’s September 2001 profile confirms via former Cabot senior tutor Julian Chang: “It was on the third floor, hanging outside his window.”
That same profile records the intellectual journey that followed. Irritated that his broker had paid him below the intrinsic value of an option — “I had been arbed,” Griffin told Institutional Investor, “And I took it upon myself to find out why” — he went to the Harvard Business School library. The Institutional Investor profile records: “Camping out at the Harvard Business School library, Griffin spent hundreds of hours imbibing finance theory from the capital asset pricing model to the Black-Scholes options pricing model.” He soon built his own convertible bond pricing model. The same profile, published in 2001, notes that Citadel was then using version 600 of it — indicating that by 2001, the model had already been iterated hundreds of times, a process that has continued since.
Before Black Monday — October 19, 1987, when the Dow fell 22.6% in a single session — Griffin was short. His model had identified that embedded options in certain convertibles were overpriced relative to the underlying equities. He had $265,000 at risk. In his Risk.net lifetime achievement interview, Griffin reflected: “it’s what we do before the event, because once it starts it’s like a bolt of lightning.”
The documented result: Citadel’s founding fund, Wellington Partners, generated 10-year net annual returns of 30.01% as of 2001 — among the best records in the industry at the time. In its first two full years, Citadel returned 43% in 1991 and 40% in 1992, trading convertible bonds in U.S. and Japanese markets. Today, Citadel explicitly runs volatility arbitrage in interest rate options, dispersion trading between index and single-stock implied vol, and cross-asset volatility arbitrage when correlation assumptions break down — each a direct monetization of a specific BS assumption failure.
The convertible arb trade structure is elegant in its logic. Buy a convertible bond trading below theoretical value; short the appropriate amount of underlying stock to hedge equity risk. The delta hedge strips out directional exposure. What remains is a long position in the embedded option’s cheapness — you are long the gap between what BS prices the embedded option at and what your proprietary model says it is worth. As the Institutional Investor profile documents, when Russia defaulted in 1998 and LTCM collapsed, Citadel had already locked up capital and de-levered ahead of the crisis, then became a rare buyer as desperate funds sold bond inventory at fire-sale prices. Their model said the embedded options were cheap. The market’s distress confirmed the gap. Citadel returned 30.5% that year.
Boaz Weinstein and Saba Capital: Capital Structure Arbitrage When Credit Spreads and Equity Implied Volatility Diverge
Where Citadel exploits mispriced embedded options within a single instrument, Saba Capital exploits a different BS blind spot entirely: the model prices equity in isolation and has no mechanism linking it to credit market signals. When those two markets diverge, one of them is wrong — and BS-calibrated equity vol cannot detect which one.
Saba Capital’s Boaz Weinstein has articulated this failure more precisely than almost any other practitioner on record. In a July 2017 interview with The Octavian Report, Weinstein described the structural disconnect:
On the volatility risk premium: “In the U.S. the phenomenon of low implied volatility really is only short-term out to three months. After that, the levels become less attractive. Eventually, out to two to three years, implied volatility is actually over double the level of recent realized volatility.”
On what creates and sustains the opportunity: “There are sellers of volatility that look at it as an attractive carry trade since the credit market offers so little carry at present. Others, such as volatility control funds, sell volatility as it goes lower to keep a constant amount of exposure.” The systematic selling by these structural participants — not rational valuation — keeps IV persistently above RV.
On why the reversal, when it comes, is violent: “if there is one, it will be much more severe because you have all of this short interest.”
Merton’s structural model — which treats equity as a call option on the firm’s assets, with credit spreads reflecting default probability — implies what equity implied vol should be given observed credit spreads. When the two diverge materially, one of them is mispriced. BS-calibrated equity vol cannot detect this divergence. Merton’s framework can.
In early 2020, Saba had built positions exploiting exactly this disconnect. According to Risk.net’s account of Saba’s Hedge Fund of the Year win, the fund had added approximately $1 billion to its credit curve-flattening trade and roughly $500 million to its short portfolio of lower-quality CDS — names in travel, retail, and energy where credit spreads implied materially higher stress than BS-calibrated equity vol was pricing. When the pandemic repricing hit in March, credit spreads moved violently toward where Saba’s capital structure model had priced them. The flagship Capital Master Fund gained 25.5% in January–February 2020, and Saba’s dedicated Tail Fund returned 99% in March 2020 alone. The alpha source was the gap between what BS-calibrated equity vol implied and what Merton’s credit model implied for forward default probability.
Mark Spitznagel and Universa: Systematically Buying the Tail Risk Black-Scholes Prices as Nearly Worthless
Both Citadel and Saba exploit specific mispricings that require a cross-instrument or cross-market model to detect. Universa takes a conceptually different position: it argues that the foundational distributional assumption of BS — log-normal returns — is structurally wrong in the direction that matters most, and that the market can never fully correct for this during extended calm periods.
In his April 7, 2020 investor letter — reported by the Wall Street Journal and confirmed by Yahoo Finance — Spitznagel explained the mathematical basis: “the big losses are essentially ALL that matter to your rate of compounding, not the small losses — and not even the big or small gains. The big losses literally destroy your geometric returns and, equivalently, your wealth, through what I have called the ‘volatility tax.’ For risk mitigation to be effective, it therefore must focus primarily on mitigating those big, rare losses.”
The trade structure: approximately 3.3% of a reference portfolio in deep out-of-the-money put options on the S&P 500 and financial companies. The remaining 96.7% in equities. Deep OTM puts are priced cheaply by BS precisely because its log-normal distribution assigns low probability to crashes. Spitznagel buys them systematically during low-vol regimes when they are cheapest — because he believes the true fat-tailed distribution assigns materially higher probability to tail events than BS acknowledges.
A critical technical note on Universa’s reported returns, essential for accurate interpretation: The figures reported are returns on required invested capital — the small options allocation — not on total portfolio AUM. The options position itself showed a return of approximately 3,612% in March 2020 and 4,144% for Q1 2020 on the allocated premium, as confirmed by Bloomberg’s reporting on the investor letter. On a full reference portfolio basis (3.3% Universa + 96.7% S&P 500), Universa documented +0.4% in March 2020 versus −12.4% for a pure S&P 500 portfolio — that is the real-world portfolio impact.
The strategy repeated in April 2025: Universa posted approximately 100% return on capital amid tariff-driven volatility, confirmed to Reuters via a fund allocator. Spitznagel declined to confirm the figure, telling Reuters he sees markets remaining in a temporary “Goldilocks zone” before further turbulence. The persistence of this result — 2008, 2020, 2025 — demonstrates that deep OTM puts remain structurally mispriced during complacent periods. No amount of generic VRP arbitrage fully corrects the log-normal distribution’s underestimation of crash probability.
Dispersion Trading and the Correlation Risk Premium: The Dimension Black-Scholes Doesn’t Model
Beyond individual instrument mispricing and cross-market divergence, there is a third structural dimension of BS failure: the model prices each option in isolation, with no mechanism for the correlation between assets. This gap creates the Correlation Risk Premium — one of the most systematically traded structural edges in institutional options markets today.
Index option pricing depends critically on correlation, because the variance of an index portfolio equals the weighted sum of individual variances plus all pairwise covariance terms:
σ²_Index = Σᵢ Σⱼ wᵢ · wⱼ · σᵢ · σⱼ · ρᵢⱼWhen you compare the implied volatility of an index option against the weighted average implied vol of its constituent single-stock options, you can recover implied correlation — what the market is pricing for how correlated the constituent stocks will be. Research using S&P 500 data from 1996–2003 documented implied correlation running an average of 18 percentage points above realized correlation. The gap is the Correlation Risk Premium (CRP) — structural because institutional investors systematically over-purchase index options for portfolio hedging, inflating index-level implied vol relative to single-stock vol.
The trade that monetizes this is dispersion: sell index options (overpriced by the CRP); buy individual constituent options (fairly priced). You are short implied correlation. When stocks move independently — as they do in stock-specific fundamental environments — the position profits. When macro shocks send all stocks crashing together, correlation spikes and the position loses.
A rigorous academic backtest of S&P 500 dispersion strategies from 2000–2017, published in MDPI Mathematics, found returns of 14.52% and 26.51% per annum after transaction costs, with Sharpe ratios of 0.40 and 0.34 across two different weighting methodologies. These are out-of-sample measurements over a 17-year period, not theoretical projections.
The execution barrier is unforgiving: a worked practitioner example from Interactive Brokers shows a theoretical variance gap of 0.00503 vol² points being entirely consumed by basket crossing costs of approximately 300 basis points. The firms that survive in dispersion — Citadel, SIG, Jane Street, Optiver — do so because their execution infrastructure allows them to trade constituent baskets at spreads competitors cannot match.
Even as April 2025 correlation spiked to two-year highs during tariff volatility, Hedgeweek reported in May 2025 that sophisticated managers kept dispersion trades profitable by constructing focused baskets of names with elevated realized volatility. BBVA flow derivatives strategist Michalis Onisiforou confirmed: despite the correlation spike, “dispersion trades have been profitable over the last few months.”
The Alpha Decay: Why Generic Short-Volatility Stopped Working After 2010
The strategies documented above share one characteristic: they require a specific, proprietary analytical edge to execute. That specificity is not accidental — it is a direct response to the fact that the generic form of the VRP trade has been competed away.
Ian Dew-Becker and Stefano Giglio’s paper documents the timeline precisely: the CAPM alpha of traded delta-hedged options on the S&P 500 — historically strongly negative, meaning buyers consistently underperformed the CAPM benchmark — broke somewhere around 2010 and has since converged to zero, well before COVID. The paper’s own abstract states it plainly: “over the past 15 years, option alphas have become indistinguishable from zero.” The compression was not caused by a crisis. It was competed away.
The mechanism: as retail brokers eliminated commissions on options trading and vol-strategy funds proliferated, the supply of option-selling capacity increased dramatically. More sellers chasing the same premium compressed the IV-RV gap toward a fair compensation for volatility risk — eliminating the excess return above that fair compensation.
The direct implications for current strategy are threefold. Raw short-vol positions — selling ATM straddles and collecting the generic VRP — no longer generate the same alpha above CAPM they did pre-2010; the excess return has been largely competed away. Specific signal advantage remains viable: funds that can forecast which specific options are mispriced — through earnings-vol analysis, credit-equity arbitrage, cross-asset correlation signals — still find edge because their signal is proprietary and not subject to the same crowding. Citadel’s documented strategies — earnings volatility, rate options term structure, commodity options — are specific-signal trades, not generic VRP harvests. And deep tail mispricing persists: Universa’s continued success demonstrates that no amount of generic vol-selling arbitrage fully corrects the log-normal distribution’s structural underestimation of crash probability during extended calm regimes.
The Volatility Smile and Skew: Black-Scholes’ Most Visible and Most Continuously Traded Failure
The alpha decay of generic short-vol does not mean options surfaces have become efficiently priced — it means the average level of options has been competed toward fair value. The shape of the vol surface — the smile and skew — remains persistently anomalous, and remains continuously traded.
If Black-Scholes were exactly correct, all options on the same underlying with the same expiry would carry the same implied volatility. In practice, out-of-the-money puts consistently carry higher implied vols than at-the-money options — the volatility skew — while OTM calls carry lower vols. Hull and White document that if implied volatilities differ systematically by strike and are treated as independent of the asset price, arbitrage opportunities must exist.
The skew is the market’s built-in correction for the fact that equity returns have negative skewness and excess kurtosis — crashes happen more frequently and more severely than the normal distribution predicts. Sophisticated funds read the vol surface as a real-time diagnostic: where is BS most wrong? Where is the skew too steep (puts overpriced) or too flat (puts underpriced)? Citadel explicitly runs term structure trades betting on volatility mean reversion and cross-asset vol arbitrage when correlation assumptions break down — both direct readings of the vol surface for local mispricings that constant-volatility BS cannot capture.
📊 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
The Synthesis: Three Strategies, One Framework, Five Decades of Evidence
Citadel, Universa, and Saba operate from a single conceptual framework: Black-Scholes is a quoting language, not a pricing truth. Haug and Taleb establish this academically. Griffin, Spitznagel, and Weinstein confirm it in practice — each with a distinct analytical model that differs from BS in precisely the dimension their market exploits.
Citadel: a convertible bond pricing model that was BS in 1987, iterated through hundreds of versions against market reality — each improving on how the prior one mispriced embedded options. Universa: a fat-tail distribution model that says deep OTM puts are worth more than BS implies, buying them precisely when BS says they are cheapest, and waiting. Saba: a capital structure model that links credit default probabilities to equity implied vol in ways that BS — which treats equity in isolation — never does.
The call formula C = S·N(d₁) − K·e^(−rT)·N(d₂) assigns a price by assuming σ is constant and returns are normally distributed. Each of these funds makes money by knowing — with evidence, with tested models, with documented risk management — that σ is not constant and that returns are not normally distributed. The gap between what the formula prices and what a better model prices is the alpha. Five decades of documented returns across the most profitable trading firms in history confirm that the gap is real, persistent, and large enough to build multi-billion-dollar franchises on.
Verified Primary Sources
Haug & Taleb — “Option Traders Use (very) Sophisticated Heuristics, Never the Black-Scholes-Merton Formula” (JEBO, 2011): https://www.researchgate.net/publication/223890470_Option_Traders_Use_very_Sophisticated_Heuristics_Never_the_Black-Scholes-Merton_Formula
Bakshi & Kapadia — “Delta-Hedged Gains and the Negative Market Volatility Risk Premium” (RFS, 2003, Vol. 16, №2, pp. 527–566): https://academic.oup.com/rfs/article-abstract/16/2/527/1579962
Carr & Wu — “Variance Risk Premia” (RFS, 2009): https://engineering.nyu.edu/sites/default/files/2019-01/CarrReviewofFinStudiesMarch2009-a.pdf
Dew-Becker & Giglio — “The Decline of the Variance Risk Premium” (CAPM alpha of traded options, post-2010 convergence to zero): https://www.dew-becker.org/documents/synth_opt.pdf
Ken Griffin — Stanford GSB Interview: https://www.gsb.stanford.edu/insights/ken-griffin-investing-winning-why-hes-focused-future
Institutional Investor — “Boy Wonder” Griffin Profile (September 2001 — primary source for satellite dish location, “hundreds of hours,” Wellington 30.01% returns, version 600): https://www.institutionalinvestor.com/article/2btfmc4i914x7pya9zwg0/home/boy-wonder
Ken Griffin — Risk.net Lifetime Achievement Interview: https://www.risk.net/awards/7755351/lifetime-achievement-award-ken-griffin
Citadel Trading Strategies — DayTrading.com: https://www.daytrading.com/citadel-ken-griffin-strategies
HedgeVision — Citadel 1991/1992 returns:
Boaz Weinstein — Octavian Report Interview (2017):
Saba Capital +25.5% (Jan–Feb 2020) — Bloomberg: https://www.bloomberg.com/news/articles/2020-03-06/boaz-weinstein-thrives-in-market-chaos-with-a-25-5-gain-in-2020
Saba Capital Hedge Fund of the Year — Risk.net ($1B flattener, $500M short CDS, Tail Fund +99% March 2020): https://www.risk.net/awards/7738506/hedge-fund-of-the-year-saba-capital-management
Universa +4,144% Q1 2020 — Yahoo Finance / Wall Street Journal: https://finance.yahoo.com/news/universa-investments-march-performance-164113528.html
Universa Investor Letter Analysis (Spitznagel “volatility tax” quote, full-portfolio +0.4% vs. S&P −12.4%): https://www.linkedin.com/pulse/tail-risk-hedging-perpetual-profitability-how-universa-george-p-babu
Universa +100% April 2025 — Hedgeweek: https://www.hedgeweek.com/black-swan-hedge-fund-universa-up-100-amid-april-volatility-says-allocator/
Dispersion Trading Backtest — MDPI Mathematics (2000–2017, 14.52%/26.51% p.a. after costs): https://www.mdpi.com/2227-7390/8/9/1627
Correlation Risk Premium / DSPX — Resonanz Capital: https://resonanzcapital.com/insights/dispersion-trading-and-the-dspx-index
Dispersion Trading in Practice — Interactive Brokers: https://www.interactivebrokers.com/campus/ibkr-quant-news/dispersion-trading-in-practice-the-dirty-version/
Hedge Funds Refine Dispersion Trades — Hedgeweek (May 2025): https://www.hedgeweek.com/hedge-funds-refine-dispersion-trades-amid-market-volatility-shift/
Hull & White — Volatility Surfaces paper: https://www-2.rotman.utoronto.ca/~hull/DownloadablePublications/DHSPaperdraft7.pdf
Volatility Arbitrage — Wikipedia: https://en.wikipedia.org/wiki/Volatility_arbitrage
Implied vs. Realized Volatility — MenthorQ: https://menthorq.com/guide/implied-vs-realized-volatility/
Editorial fact-check notes for transparency:
Satellite dish: The Institutional Investor September 2001 profile, fetched directly, says: “It was on the third floor, hanging outside his window.” Not the roof.
Griffin’s age: The II article calls him “the 18-year-old” in the relevant fall 1987 context. He turned 19 on October 15, 1987 — days before Black Monday itself.
“Hundreds of hours”: Confirmed verbatim in the II article: “Griffin spent hundreds of hours imbibing finance theory.” Attribution stands.
Bakshi & Kapadia data period: Full sample = January 1988 to December 1995 (two subsamples). The “1991” start date belongs to a separate Bakshi-Kapadia Journal of Derivatives paper on individual equity options, not this RFS paper.
Dew-Becker & Giglio metric: The paper’s primary metric is CAPM alpha. The term “information ratio” appears once in the paper — in a narrow technical sentence about confidence bands on synthetic options — but the decline narrative throughout maps to CAPM alpha. Article updated accordingly.
Universa returns: All returns cited (3,612%, 4,144%, 100%) are returns on invested capital in the options allocation, not on total AUM. Full-portfolio context provided.
Saba’s 2020 strategy: Primary instrument was capital structure arbitrage via credit default swaps — confirmed by the Risk.net Hedge Fund of the Year account. Not vanilla equity puts.
About the Author
Navnoor Bawa researches quantitative finance, derivatives pricing, and institutional trading strategy.
If this article added value to your thinking, you can follow the work across these platforms:
📺 YouTube — The Mathematical Trader: In-depth video breakdowns of quantitative strategies, derivatives mechanics, and institutional market structure → youtube.com/@TheMathematicalTrader
💼 LinkedIn: Research updates, article announcements, and professional discussion → linkedin.com/in/navnoorbawa
📊 Patreon — Exclusive Quantitative Research: Deep-dive research, institutional-grade trading strategies, and analysis not published anywhere else → Join here
Cover photograph: Massachusetts Institute of Technology, CC BY-SA 4.0, via Wikimedia Commons.