Kelly Criterion in Trading: Optimal Bet Size
The Kelly Criterion calculates the mathematically optimal bet size for maximizing long-term growth, but full Kelly is too volatile for most traders.
Kelly Criterion in Trading: Optimal Bet Size
Bet too small and you leave growth on the table. Bet too big and a losing streak ends you. Kelly finds the line in between.
The Kelly Criterion is a formula developed by John Kelly in 1956 that calculates the fraction of your bankroll to risk in order to maximize long-term compound growth. Originally built for gambling, it translates directly to trading.
The formula
Kelly% = W − (1 − W) / R
Where:
- W = win rate (as a decimal, e.g., 0.55)
- R = average win ÷ average loss (reward-to-risk ratio)
Example: Win rate 55%, average win $300, average loss $100 (R = 3).
Kelly% = 0.55 − (0.45 ÷ 3) = 0.55 − 0.15 = 0.40
Full Kelly says risk 40% of your account per trade. That is insane — read on.
Why full Kelly is dangerous
Full Kelly maximizes expected logarithmic growth, but the path is brutal:
| Kelly fraction | Long-term growth | Drawdown depth |
|---|---|---|
| 100% (full) | Maximum | Extreme (50%+ common) |
| 50% (half) | ~75% of max | Moderate |
| 25% (quarter) | ~50% of max | Mild |
A 40% per-trade risk means a 5-trade losing streak leaves you with less than 8% of your account. Few traders can stomach that.
The practical answer: fractional Kelly
Most professional traders use half-Kelly or quarter-Kelly:
- Half-Kelly = Kelly% ÷ 2
- Quarter-Kelly = Kelly% ÷ 4
For the example above, half-Kelly is 20%, still high for trading. Quarter-Kelly is 10%. Many discretionary traders settle between 1%–2% total risk — well below Kelly — because their edge estimates are unreliable.
Estimating your inputs honestly
Kelly is only as good as your inputs. Three traps:
- Overstating win rate: 100 backtested trades isn't enough to trust 60% as truth
- Ignoring slippage and fees: They shrink R more than you think
- Non-stationary edge: A strategy that worked in 2021 may not work in 2026
If your true win rate is 50% (not 55%) and R is 2 (not 3), Kelly drops from 40% to 0%. Trade a non-edge with Kelly sizing and you go broke optimally.
How to actually use Kelly
- Backtest 200+ trades to estimate W and R conservatively
- Compute full Kelly, then divide by 4 (quarter-Kelly)
- Cap the result at 1%–2% of account
- Cross-check with the position size calculator
- Track results in a journal and re-estimate W and R quarterly
Summary
Kelly tells you the mathematically optimal size — not the psychologically survivable one. Use fractional Kelly as a ceiling, never as a target. If your computed Kelly exceeds 2%, treat it as a warning that your edge estimate is probably too optimistic, not as permission to size up.