The Kelly criterion gives the fraction of capital to commit to a favourable bet or investment in order to maximise long-run growth. It balances the gain from a good edge against the risk of ruin from betting too much.
The formulas
- For a simple bet: f* = p - q / b, where p is the chance of winning, q = 1 - p, and b is the amount won per unit staked.
- For an investment with continuous returns: f* is roughly the expected excess return divided by the variance of returns, or (return - risk-free rate) / volatility squared.
Two worked examples
| Case | Inputs | Result |
|---|---|---|
| Simple bet | Win 55% of the time, win or lose equal amounts (b = 1) | f* = 0.55 - 0.45 / 1 = 0.10, or 10% of capital |
| Investment | Excess return 6%, volatility 18% | f* = 0.06 / 0.182 = 0.06 / 0.0324, about 1.85, or 185% of capital |
The second result means the formula would call for borrowing, which shows why it is rarely used at full size. The inputs are illustrative.
Why practitioners use less than full Kelly
- Estimates are uncertain. The formula assumes the edge is known exactly. If the true edge is smaller, full Kelly bets too much, and over-betting reduces growth and can cause deep losses.
- Drawdowns are large. Full Kelly can produce falls most investors would not tolerate. See drawdown.
- Fractions are common. Half Kelly, for example, gives up a modest share of growth for much smaller swings.
Where it connects
Position sizing ties in with risk management, volatility targeting and portfolio optimization. It is a sizing idea, not a way to find an edge, and it relies on inputs that must be estimated.
Frequently asked questions
What is the Kelly criterion?
It is a formula for the fraction of capital to stake on a favourable bet to maximise the long-run growth rate of capital.
Why do investors use half Kelly?
Because the inputs are estimated with error and full Kelly produces large swings. Using a fraction reduces the risk of over-betting at a modest cost to growth.
Can the Kelly criterion suggest leverage?
Yes. When the edge is large compared with the risk, the formula can call for more than 100% of capital, which is one reason it is usually applied with caution.
Does the Kelly criterion guarantee profit?
No. It maximises expected long-run growth under its assumptions, but results in any period can be poor, and it depends on knowing the true odds.