KellyCriterion<T>
Models & Types · AiDotNet.Finance.Portfolio
Kelly-criterion position sizing — the bet fraction that maximizes long-run log-growth of capital.
Implements: IPositionSizer<T>
For Beginners
Kelly tells you what fraction of your capital to put on a trade to grow wealth fastest over many trades. Bet more than Kelly and you risk ruin; bet a fraction of it (“half-Kelly”) for much lower volatility at a small growth cost. A negative Kelly means “no edge — don’t take the trade.”
How It Works
AiDotNet’s portfolio optimizers cover weight allocation (mean-variance, HRP, Black-Litterman) but not Kelly bet sizing. This fills that gap with the two standard forms — discrete (win probability + payoff odds) and continuous (mean / variance of returns) — plus fractional Kelly, which is what practitioners actually use because full Kelly is famously over-aggressive.
Constructors
| Constructor | Summary |
|---|---|
KellyCriterion() |
Properties
| Property | Summary |
|---|---|
IPositionSizer<T> Default { get; } |
Shared stateless default instance for injection as an IPositionSizer. |
Methods
| Method | Summary |
|---|---|
T Continuous(T expectedReturn, T variance) |
Continuous Kelly fraction for (approximately) Gaussian returns: f* = μ / σ², where expectedReturn is μ and variance is σ². |
T Discrete(T winProbability, T winLossRatio) |
Discrete Kelly fraction: f* = p − (1 − p) / b, where winProbability is p and winLossRatio is b (net payoff on a win per unit risked on a loss). |
T Fractional(T kellyFraction, double fraction) |
Fractional Kelly: kellyFraction scaled by fraction (e.g. |
T FromReturns(IEnumerable<T> returns, double fraction) |
Continuous Kelly estimated from a return series: computes the sample mean and (population) variance of returns and applies Continuous. |