Kelly Criterion for Position Sizing Explained
The Kelly Criterion calculates the mathematically optimal bet size from your win rate and payoff ratio. Here's the formula, a worked example, and why traders use half-Kelly.
The Kelly Criterion is a formula that calculates the fraction of your capital to risk on a given bet or trade to maximize long-term growth, based on your win probability and payoff ratio. Originally developed for information theory at Bell Labs in 1956, it was later adopted by gamblers and then traders as a mathematically grounded alternative to guessing at position size.
The Formula
Kelly % = W - [(1 - W) / R]
Where:
- W = your win rate (probability of a winning trade, as a decimal)
- R = your win/loss ratio (average winning trade size ÷ average losing trade size)
The output is the percentage of your capital to risk on the next trade.
Worked Example
Suppose your trading history shows:
- Win rate (W) = 45% (0.45)
- Average win = $300, average loss = $150, so R = 300/150 = 2.0
Kelly % = 0.45 - [(1 - 0.45) / 2.0] = 0.45 - (0.55 / 2.0) = 0.45 - 0.275 = 0.175
Full Kelly says to risk 17.5% of your capital on this trade. For most traders, this number is uncomfortably large, which brings up the most important practical caveat to the formula.
Why Full Kelly Is Rarely Used in Practice
| Issue | Why It Matters |
|---|---|
| Kelly assumes you know your true win rate and payoff ratio exactly | In real trading, both are estimates from limited historical data and will drift over time |
| Full Kelly sizing produces large drawdowns | Even with a genuine edge, full Kelly can produce 50%+ equity swings, which is psychologically and practically unworkable for most traders |
| Overestimating your edge is catastrophic | If your real edge is smaller than your estimate, full Kelly sizing accelerates account damage rather than growth |
Because of this, most practitioners use half-Kelly or quarter-Kelly — simply taking 50% or 25% of the calculated Kelly percentage. This sacrifices some theoretical long-term growth rate in exchange for dramatically smoother equity curves and much more tolerance for estimation error in your inputs.
| Sizing Approach | % of Capital Risked (from example above) |
|---|---|
| Full Kelly | 17.5% |
| Half-Kelly | 8.75% |
| Quarter-Kelly | 4.375% |
Why Kelly Can Go Negative
If your edge is negative — meaning W - [(1-W)/R] computes to a negative number — the formula is telling you not to take the trade at all, or that the strategy has no statistical edge as measured. A negative Kelly output is a genuinely useful signal: it means no position size makes this trade a good idea, and the fix is to improve the strategy's win rate or payoff ratio, not to adjust the position size.
What Kelly Requires You to Actually Know
Kelly sizing is only as good as the inputs. To use it responsibly you need:
- A statistically meaningful sample size of past trades for the specific strategy (see How Many Trades Do You Need for a Statistically Valid Backtest? for how many trades that actually requires)
- A stable edge — Kelly assumes the win rate and payoff ratio you calculated will hold going forward, which is a much stronger assumption in a live, changing market than in a casino game with fixed odds
- Willingness to recalculate regularly as your live results update your win rate and payoff ratio estimates
Kelly vs Fixed-Percentage Risk
Most retail trading education teaches a simpler rule: risk a fixed 1-2% of account equity per trade, regardless of the specific setup's edge. Kelly is more precise in theory (it scales position size to the actual edge of each setup) but more fragile in practice (it's highly sensitive to bad input estimates). A reasonable middle ground many traders use: apply fixed fractional risk as a hard ceiling, then use half-Kelly logic as a way to scale size down for lower-conviction setups relative to higher-conviction ones, rather than using raw Kelly as an absolute sizing formula.
Summary
The Kelly Criterion converts your win rate and payoff ratio into a mathematically optimal position size for maximizing long-term capital growth, but full Kelly's aggressive output and sensitivity to bad estimates mean most traders should use half- or quarter-Kelly in practice. Its real value isn't the exact percentage it spits out — it's the discipline of quantifying your edge before sizing a position, and treating a negative Kelly output as a signal to skip the trade entirely.
Related reading:
We're Cooking Something Great.
Revealing Soon.
TradeThesis is being rebuilt from the ground up. The 5-agent AI research pipeline is coming back sharper than before.
No sign-up needed. Just watch this space.