Kelly Criterion

Kelly Criterion is the repeated-bet sizing frame in E153.股神的牌局：复利公式 + 凯利公式. The episode presents it as a tool for maximizing long-term logarithmic capital growth while avoiding ruin, not as a license to concentrate aggressively on a single attractive trade.

E144.交易的艺术：不预测，统计优势，分散红利，随机波动 uses the same repeated-game logic in a trend-trading context. It discusses win rate, payoff ratio, and small signal-combination backtests as inputs one might use to estimate a Kelly-like position, while warning that past signal statistics should not be treated as future certainty.

Key Claims

• Kelly applies best to repeated or effectively infinite games where survival matters more than one dramatic win.
• The formula requires estimates of win probability and payoff ratio, but the episode stresses that real investors often overestimate both.
• When no edge exists, the Kelly logic says not to play, or to avoid long exposure, rather than force a trade.
• Fractional Kelly, such as half Kelly or quarter Kelly, reduces theoretical maximum return but improves drawdown tolerance and emotional stability.
• The episode’s implementation advice is to review one’s own trade records by strategy or asset class, estimate win rate and payoff ratio, then discount the result.
• E144 reinforces that a low hit rate can still support a position if payoff asymmetry is high enough, but only if the estimates come from comparable repeated trades.
• High leverage can make a positive-expectation setup fragile because a single bad path can remove the investor from the repeated game.
• Add-on buying requires higher confidence or payoff than the first entry because the position is already exposed to correlated risk.

Connections

• Edward Thorp and Claude Shannon — historical and theoretical figures used to explain the idea.
• Position Sizing — practical application layer for trades and portfolios.
• Investment Edge — the criterion matters only when the bettor has positive expectation.
• Compounding Growth Formula — Kelly governs the position-size part of the broader growth frame.
• Investment Risk Management, Stop-Loss Discipline, and Pyramiding — trading discipline concepts reinforced by the source.
• No-Prediction Trading — E144’s trend-system use case for repeated-bet sizing.