Kelly Criterion
Formula Details
1. Discrete Return Version
Parameter description: $p$ is the win probability, $q$ is the loss probability, $b$ is the odds of "how much you net gain from one win," and $f^*$ is the fraction of capital that should be invested each time.
Detailed explanation: This formula seeks the bet size that maximizes the long-term compound growth rate. It is not about maximizing a single win, but about making capital grow fastest over many repeated games.
2. Expected Growth Interpretation
Parameter description: $g(f)$ is the long-term logarithmic growth rate when the bet fraction is $f$.
Detailed explanation: The Kelly criterion is essentially maximizing "long-term average logarithmic wealth growth," so it is naturally biased toward compounding rather than maximizing single-round returns.
3. Continuous Return Version in Investment Scenarios
Parameter description: $\mu$ is the asset's expected continuous return, $r$ is the risk-free rate, and $\sigma$ is the return volatility.
Detailed explanation: In investment scenarios, if returns are treated as a continuous compounding process, the Kelly fraction can be written as "excess return / variance." This is also a common continuous-time approximation in asset allocation and position management.
Detailed Explanation of the Kelly Criterion
What Problem Does It Solve?
Why Is It Important?
Application Scenarios
• Asset allocation: when different assets have clear risk-return characteristics, calculate their respective theoretical optimal weights.
• Gambling/competition: in games with known odds such as blackjack and poker, professional players use Kelly to manage funds.
• Venture investment: estimate capital allocation across projects with different success probabilities.
Deficiencies and Limitations
• Full Kelly has extremely high volatility, and account drawdowns can exceed 50%, which most people cannot psychologically withstand, so in practice "half Kelly" or more conservative fractions are often used.
• It assumes each game is independent and parameters are constant, whereas financial markets have complex features such as correlation and fat tails.
• If you overestimate your win-rate edge, using the Kelly formula will instead accelerate losses.
Theoretical Sources / Further Reading
• Edward O. Thorp: Beat the Dealer
• Edward O. Thorp: Beat the Market
• MacLean, Thorp, Ziemba: The Kelly Capital Growth Investment Criterion