Confirmed claims below survived 3-vote verification. Synthesis interrupted by usage limit.
Benter's Hong Kong betting operation was profitable in four of its first five seasons, with the single losing season costing about 20% of starting capital, and the cumulative log-wealth graph shows roughly e^3.5-e^4 growth over ~2,500 races bet. https://gwern.net/doc/statistics/decision/1994-benter.pdf | 3-0
Building the edge required roughly ten man-years of effort: five man-years to build the database and a model with significant advantage, plus five more to reach high profitability — supporting the thesis that the niche was too effortful for casual competitors. https://gwern.net/doc/statistics/decision/1994-benter.pdf | 3-0
The core technical insight Benter published is that a raw fundamental model is biased relative to the public odds and must be combined with the public's implied probabilities in a second-stage logit (his combined model raised pseudo-R² from .1218 public / .1245 fundamental to .1396 combined, ΔR² = .0178, which he says indicates significant profits). https://gwern.net/doc/statistics/decision/1994-benter.pdf | 3-0
Benter explicitly used fractional Kelly sizing (e.g. 1/2 or 1/3 Kelly) because full Kelly is unsafe when the advantage estimate itself has error — overestimating advantage by a factor of two makes full-Kelly betting negative-growth — and his operation employed a conservative fractional Kelly throughout. https://gwern.net/doc/statistics/decision/1994-benter.pdf | 3-0
Benter quantified the edge's capacity ceiling: maximum expected profit is about 0.25-0.5% of total per-race pool turnover for an effectively infinite bankroll (winning 0.1-0.2% of track turnover being a realistic start-up goal), which is why he targeted Hong Kong's extremely large pools (> USD 10,000,000 per race, ~19% average track take) and why small-pool tracks cannot sustain a professional operation. https://gwern.net/doc/statistics/decision/1994-benter.pdf | 3-0
Thorp's method for finding and exploiting edges followed an explicit three-step protocol — idea, quantitative development, real-world verification — which he validated cheaply: his first casino trip in 1961 turned a $10,000 bankroll into an $11,000 win in 20 hours of full-scale betting, close to his forecast midpoint of $10,000. https://www.edwardothorp.com/wp-content/uploads/2016/11/thorpwilmottqfinrev2003.pdf | 3-0
Thorp's documented sizing discipline was fractional Kelly with a staged ramp-up: he cut his backers' proposed $100K bankroll to $10K as a safety measure, laddered bet sizes from $1-10 up to $50-500 over roughly 15 hours of verification play, bet fractional Kelly within each range, and only bet full Kelly at full scale. https://www.edwardothorp.com/wp-content/uploads/2016/11/thorpwilmottqfinrev2003.pdf | 3-0
Thorp explicitly sized each edge's capacity in dollars before committing: he valued blackjack at $1 million to him personally and convertible hedging at $100 million; Princeton Newport Partners ultimately made about $250 million for its partners, half or more from convertibles. https://www.edwardothorp.com/wp-content/uploads/2016/11/thorpwilmottqfinrev2003.pdf | 3-0
Thorp names consistent overbetting (violating Kelly) as the ruin mechanism behind documented blow-ups, attributing the misfortunes of Victor Niederhoffer and LTCM in part to it, and notes LTCM also fell to the fallacies of trusting historical limits and lognormal tail models. https://www.edwardothorp.com/wp-content/uploads/2016/11/thorpwilmottqfinrev2003.pdf | 3-0
Thorp documents both his idea-search doctrine and edge decay: ideas came from sitting and thinking, academic journals, reading, and networking; his statistical arbitrage (short-term reversal found Dec 1979/Jan 1980, ~20% annualized pre-cost) compounded 26%/yr net before fees Aug 1992-Oct 2002 (~$350M profit), and in all three of his examples the inefficiency diminished only gradually over many years, requiring continuous R&D. https://www.edwardothorp.com/wp-content/uploads/2016/11/thorpwilmottqfinrev2003.pdf | 3-0
Thorp independently derived a risk-neutral option/warrant pricing formula years before Black-Scholes was published (1973), and deliberately kept it private as a proprietary trading tool for his warrant hedges rather than publishing it. https://www.aqr.com/-/media/AQR/Documents/Insights/Interviews/AQR-Words-from-the-Wise-Ed-Thorp.pdf | 3-0
Thorp's hedge fund strategies (Princeton Newport Partners, later Ridgeline Partners) were profitable every single year for 37 years, from 1966 to 2002. https://www.aqr.com/-/media/AQR/Documents/Insights/Interviews/AQR-Words-from-the-Wise-Ed-Thorp.pdf | 3-0
Thorp documents concrete edge decay in his statistical arbitrage operation: the Bamberger industry-grouped reversal model (run as a joint business 1985-1988) faded from ~25% annualized to ~15%, and his team responded by migrating to a new factor-neutralized version using principal-component risk factors with 100-200 stocks per side capped at ~1.5% per name. https://www.aqr.com/-/media/AQR/Documents/Insights/Interviews/AQR-Words-from-the-Wise-Ed-Thorp.pdf | 3-0
Thorp's stated validation protocol for any claimed edge is that it must be logically demonstrable and survive adversarial challenge; he also states that discovered edges decay because capital inflows move prices toward correct pricing. https://www.aqr.com/-/media/AQR/Documents/Insights/Interviews/AQR-Words-from-the-Wise-Ed-Thorp.pdf | 3-0
Burry publicly documented his stock-picking methods for years before managing outside money: he wrote for Microsoft's MSN Money under the handle 'Value Doc' starting in 1997, while still training in medicine. https://substack.com/@michaeljburry/note/c-191040794 | 3-0
Burry's public track record directly seeded his fund launch: he left medicine on June 30, 2000 to found Scion Capital, and Joel Greenblatt's Gotham Capital plus White Mountains Insurance Group bought stakes in the firm — a documented link in the small-capital special-situations lineage (Greenblatt -> Burry). https://substack.com/@michaeljburry/note/c-191040794 | 3-0
His documented early operating method was concentrated deep-value in out-of-favor names: holding 12-18 stocks across depressed industries and buying within 10-15% of a 52-week low that shows price support. https://substack.com/@michaeljburry/note/c-191040794 | 3-0
Kelly strategies are dangerously sensitive to estimation error, with errors in mean estimates roughly 20x as costly as covariance errors, rising to about 100:3:1 for near-zero-risk-aversion log investors — the theoretical basis for fractional Kelly under estimation error. https://webhomes.maths.ed.ac.uk/mckinnon/blackouts/StochOptFinanceAndEnergySpringer/Chap1_KellyZiemba.pdf | 3-0
Overbetting beyond full Kelly is strictly dominated: both growth and security decline for f > f*, and betting exactly twice the Kelly fraction yields a growth rate of zero (plus the risk-free rate) — the quantitative case for capping size at or below Kelly. https://webhomes.maths.ed.ac.uk/mckinnon/blackouts/StochOptFinanceAndEnergySpringer/Chap1_KellyZiemba.pdf | 3-0
The chapter attributes documented Kelly-style wealth records to named practitioners: Benter's Hong Kong racing syndicate, Thorp's Princeton-Newport, Simons's Medallion, and Keynes's King's College endowment (Keynes estimated as approximately an 80% Kelly bettor), while Buffett and Soros are characterized as full-Kelly-like concentrated investors. https://webhomes.maths.ed.ac.uk/mckinnon/blackouts/StochOptFinanceAndEnergySpringer/Chap1_KellyZiemba.pdf | 3-0
Pari-mutuel rebates convert a losing gross betting record into a net profit for volume players — a documented 2004 real-money example churned $5,000 into $1.5M of bets, lost 7% gross, and netted about $30,000 via a ~9% average rebate; rebates cut professionals' effective track take from 13-30% to roughly 10-12%. This is the documented mechanism behind rebate-economics operators like Ranogajec. https://webhomes.maths.ed.ac.uk/mckinnon/blackouts/StochOptFinanceAndEnergySpringer/Chap1_KellyZiemba.pdf | 3-0
Ranogajec testified under oath in 2008 that his syndicate profits from tote rebates ('loyalty payments') rather than winning bets: a 10% rebate turns a 5% gross betting loss into a 5% net profit, and betting at his volume depresses pari-mutuel pools below fair odds, making the rebate the entire edge. https://varenne.art/usr/documents/press/download_url/359/the-joker-meet-zeljko-ranogajec-the-australian-who-is-the-biggest-gambler-in-the-world.pdf | 3-0