This is a detailed research piece. If you find value in institutional-quality hedge fund analysis, support this work on Patreon.
Morgan Stanley’s quantitative desk extracted $50 million from pairs trading in 1987. Arbitrageurs extract roughly $40 million (estimated) from prediction market mispricings over the April 2024–April 2025 window. The core trade (exploiting temporary deviations in related securities) remains constant. Market structure determines execution mechanics, profitability, and failure modes.
The Original Architecture: Equity Pairs (1987–2007)
Genesis: Nunzio Tartaglia’s Morgan Stanley team pioneered automated pairs trading in 1985, assembling physicists and mathematicians to identify equity pairs with high correlation (NBER Working Paper 7032). When spreads diverged by two historical standard deviations, the system initiated market-neutral positions: long the underperformer, short the outperformer. The group reported $50 million profit in 1987 before disbanding in 1989.
Performance Data: Gatev, Goetzmann, and Rouwenhorst document strong historical performance for pairs trading. For the top 20 pairs, the paper reports average monthly risk-adjusted returns of 0.67% (1963–1988) declining to 0.42% (post-1989) as hedge funds crowded the strategy (Table VI). The trade generated profit from spread compression regardless of market direction. Zero beta exposure by design.
August 2007 Collapse: Khandani and Lo (2008) documented how quantitative equity market-neutral funds experienced extraordinary losses August 6–9, 2007. Several funds reported large weekly drawdowns (a prominent example reported losses in the 30% range that week). The “Unwind Hypothesis” attributes losses to forced liquidation of large, similarly positioned portfolios creating price impact cascades. Correlation breakdown wasn’t model failure. Identical positioning across competing funds created forced deleveraging.
Key Insight: Statistical arbitrage requires predictability, volatility, and dispersion. When all three compress simultaneously, correlations that held for decades break instantly.
Futures and Commodities (1990s-2010s)
Structural Innovation: Arbitrageurs extended pairs logic to commodity futures without fundamental valuation anchors. Price co-movement in energy products creates exploitable basis spreads requiring no directional views. Futures eliminated short-sale restrictions: no uptick rule, no borrow costs, immediate execution at market prices.
Capacity-Speed Tradeoff: Competition and technology compressed mean holding periods from weeks (equity pairs) to days (futures spreads). Statistical relationships decay faster than fundamental relationships, limiting signal persistence beyond 2–4 week horizons.
Cryptocurrency Arbitrage (2017-Present)
Fragmentation Advantage: Bitcoin trades on 150+ exchanges across multiple fiat pairs, creating thousands of distinct BTC prices globally. 24/7/365 trading with high volatility generates continuous price dislocations.
Spot-Futures Mechanics: Lemvi Capital and similar funds exploit persistent funding rate premiums through basis trades. Example execution: Buy 1 BTC spot at $30,000, short $30,100 XBTU23 futures (30,100 contracts at 10x leverage). At settlement, the $100 basis profit equals 0.0033 BTC ($100 USD) regardless of spot price at expiry.
Performance Divergence: Market-neutral crypto strategies achieved 14.4% YTD returns in 2024; directional funds posted -2.5%. Lemvi Capital rotated strategy allocation as opportunities compressed: cross-exchange arbitrage (2017–2019) → futures basis (2020–2022) → options arbitrage (2023-present).
Cross-Exchange Degradation: Delta-one arbitrage yielded orders-of-magnitude larger returns in early years (2017–2018, with some anecdotal accounts citing extremely high percentage returns). Market maker professionalization compressed spreads toward zero by 2024. Profitability migrated to options where Deribit dominance creates cross-venue pricing dislocations.
Prediction Markets (2024-Present)
New Primitive: Polymarket (decentralized) and Kalshi (CFTC-regulated) create parallel universes for identical events with structural barriers to convergence.
Arbitrage Structure: Research by Saguillo et al. (2025) analyzed 86 million bets across 17,218 conditions on Polymarket, documenting estimated $40 million in arbitrage profits extracted April 2024-April 2025. Two forms exist: market rebalancing arbitrage (YES+NO prices deviate from $1.00) and combinatorial arbitrage (identical markets price differently across platforms).
Cross-Platform Example (illustrative snapshot, December 2025): Community analysis documented an example where:
Polymarket: YES = 51¢ (49¢ NO implied)
Kalshi: YES = 37¢ (63¢ NO implied)
Total: 114¢ (14¢ arbitrage opportunity)
Deploy $10,000: Buy $4,300 YES on Polymarket + $5,700 NO on Kalshi = $11,000 expected payout = $1,000 profit (10% return).
Note: These are time-stamped snapshots from community reporting. Prices change minute-to-minute. Treat such examples as illustrative, not invariant.
Resolution Risk: March 2025 UMA governance incident: whale with 25% UMA voting power manipulated $7M Polymarket market resolution despite no formal agreement occurring. 2024 government shutdown: Polymarket resolved YES (incorrect), Kalshi resolved NO (correct). Cross-platform arbitrage assumes convergent settlement. Divergent oracles eliminate guaranteed profit.
Execution Barriers:
Capital efficiency: Most opportunities yield $50–500 maximum profit per condition (Saguillo et al., 2025)
Latency requirements: 5–10 minute windows during volatility events
Fee structures: Kalshi’s fees are variable (typically 0.7–3.5% based on contract probabilities). Polymarket’s fee structure has varied over time; platform docs and third-party reporting differ (some community writeups reference effective fees on net winnings). Consult current platform documentation for specifics.
Liquidity concentration: 62% of volume in mid-September 2025 concentrated in high-profile events (elections), leaving thin orderbooks elsewhere
Institutional Professionalization: By mid-September 2025, Kalshi processed $500M+ weekly, capturing 62% market share. As institutional market makers deploy capital, spread compression follows crypto’s 2016–2018 trajectory.
The Unified Pattern
Arbitrage evolution follows a four-phase cycle across asset classes:
Phase 1: Discovery. Market structure fragmentation creates opportunities. Early entrants extract 1000%+ returns (crypto 2017, prediction markets 2024). Morgan Stanley’s $50M profit (1987) came from computational arbitrage in equity markets.
Phase 2: Technology Compression. Latency arbitrage eliminates simple mispricings. Gatev returns compressed 67bp→42bp monthly. Crypto cross-exchange spreads reached zero by 2024.
Phase 3: Complexity Migration. Profitability shifts to derivatives (crypto options), multi-leg strategies, volatility/basis trades. Statistical arbitrage holding periods compressed from weeks to days.
Phase 4: Systemic Risk. Crowding creates correlation breakdown. August 2007: prominent funds reported losses in the 30% range in a single week despite sophisticated models. Prediction markets 2024–2025: divergent oracle resolution destroys “risk-free” trades.
Conclusion
Arbitrage profitability correlates with market structure fragmentation, not model sophistication. The 14¢ Polymarket-Kalshi spread (observed in December 2025 snapshots) quantifies the cost of crossing governance systems (decentralized vs. CFTC-regulated). As latency arbitrage compresses spreads, alpha migrates to resolution risk pricing: modeling governance failures, oracle manipulation, and settlement divergence. Structural barriers matter more than execution speed.
Sources
Core Research Papers:
Gatev, E., Goetzmann, W., & Rouwenhorst, K.G. (2006). “Pairs Trading: Performance of a Relative-Value Arbitrage Rule.” Review of Financial Studies, 19(3), 797–827. PDF: Wharton | DOI
Khandani, A. & Lo, A.W. (2008). “What Happened to the Quants in August 2007?” NBER Working Paper 14465. MIT PDF | NBER
Saguillo, O., Ghafouri, V., Kiffer, L., & Suarez-Tangil, G. (2025). “Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets.” arXiv:2508.03474. arXiv PDF
Industry Analysis:
Gatev, E., Goetzmann, W., & Rouwenhorst, K.G. (1999). “Pairs Trading: Performance of a Relative Value Arbitrage Rule.” NBER Working Paper 7032. NBER PDF
“The History and Evolution of Quantitative Finance (1980s).” Medium — The Financial Journal (2023). Medium
“The Evolution of Statistical Arbitrage: Rise of Alternative Data and Shorter Holding Periods.” QuantLink (2023). QuantLink
Cryptocurrency Markets:
“Lemvi: Arbitrage and Relative Value in Crypto.” The Hedge Fund Journal (2024). HFJ
“Crypto Hedge Funds Face Stunning 2024 Struggle.” CryptoRank (December 2024). CryptoRank
“How to Arbitrage with Crypto Futures and Spot.” BitMEX Blog (August 2025). BitMEX
“Hedge Fund Strategies in Cryptoland.” The Hedge Fund Journal. HFJ
“The Comprehensive Introduction to Pairs Trading.” Hudson & Thames (2023). Hudson & Thames
Prediction Markets:
“Polymarket, UMA Communities Lock Horns After $7M Ukraine Bet Resolves.” CoinDesk (March 2025). CoinDesk
“Prediction Markets Cannot Agree on the Truth.” Monad Blog (May 2025). Monad
Rodriguez, F. “Kalshi Outpaces Polymarket in Prediction Market Volume.” CoinDesk (September 2025). CoinDesk
Polymarket Trading Fees Documentation. Polymarket Docs
Data Verification (Sources checked through September 2025):
$50M (1987): NBER WP 7032, multiple historical accounts
67bp→42bp: Gatev et al. (2006), Table VI monthly risk-adjusted returns for top-20 pairs
August 2007 losses: Khandani & Lo (2008), page 3; contemporaneous WSJ reporting
$40M prediction markets: Saguillo et al. (2025), arXiv abstract quantification
14.4% market-neutral crypto: CryptoRank industry aggregates (2024)
62% Kalshi share, $500M+ weekly: CoinDesk/Dune Analytics (mid-September 2025)
86M bets, 17,218 conditions: Saguillo et al. dataset description
UMA governance incident: CoinDesk March 2025 reporting
Important notes:
Gatev returns are monthly, risk-adjusted figures for top-performing pairs
Cross-platform price examples are time-stamped snapshots; prices are ephemeral
Fee structures (especially Polymarket) have varied; consult current platform docs
Early crypto arbitrage returns are anecdotal/retrospective industry accounts
All claims cross-verified against primary sources where available. Performance figures represent historical results.
📊 Support this research: https://www.patreon.com/c/NavnoorBawa
Cover photograph: Ajay Suresh, CC BY 2.0, via Wikimedia Commons.