Most traders pick a position size out of habit — 1% risk per trade, 2% risk per trade, or whatever number felt safe the first time they read it in a book. The Kelly Criterion replaces that guess with a formula: given your actual win rate and your actual average win-to-loss ratio, it calculates the exact fraction of your account that maximizes long-term growth. The catch is that it only works if the numbers you feed it are real — which means it only works if you're journaling.
Win rate and win/loss ratio go in, an optimal risk fraction comes out — but the formula is only as accurate as the journal behind the inputs.
2
inputs the Kelly formula needs — win rate and average win/loss ratio
30–50+
closed trades typically needed before those inputs are stable enough to trust
½ Kelly
the fraction most professional sizing models actually use, not full Kelly
0%
position size the formula recommends when your edge is negative
What the Kelly Criterion Actually Calculates
Developed by Bell Labs scientist John Kelly in 1956, the Kelly Criterion answers a specific question: given a repeated bet with a known edge, what fraction of your bankroll should you risk on each one to grow it the fastest over time, without risking ruin? Applied to trading, "known edge" means your historical win rate and your average win size relative to your average loss size — both of which come directly from your closed trades.
The formula is: K% = W − [(1 − W) / R], where W is your win rate as a decimal and R is your average win divided by your average loss. The output, K%, is the percentage of your account to risk on the next trade.
A Worked Example
Say your journal shows a 45% win rate with an average win of $600 and an average loss of $300 — an R of 2.0. Plug that in: K% = 0.45 − [(1 − 0.45) / 2.0] = 0.45 − 0.275 = 0.175, or 17.5%. That is the mathematically "optimal" risk per trade for maximizing long-term growth, and it is also, for almost every trader, a wildly aggressive number to actually risk.
| Win Rate | Avg Win / Avg Loss (R) | Kelly % |
|---|---|---|
| 35% | 2.5 | 9% |
| 45% | 2.0 | 17.5% |
| 50% | 1.5 | 16.7% |
| 60% | 1.0 | 20% |
| 40% | 3.0 | 20% |
Why Full Kelly Is the Wrong Number to Actually Risk
Full Kelly assumes your win rate and R are exact and constant forever. In real trading, they're neither — they're estimates from a limited sample, and they drift as market conditions and your own execution change. Betting the full Kelly fraction against an estimate that's even slightly optimistic produces sharp, realistic-looking drawdowns; against an estimate that's meaningfully wrong, it produces the kind of drawdown that ends accounts. This is why professional sizing models almost never use full Kelly.
- Half Kelly (½K) cuts the recommended risk in half and gives up only about a quarter of the theoretical growth rate — the standard compromise used across quantitative sizing models
- Quarter Kelly (¼K) is more common for traders whose edge estimate is based on a smaller sample or a strategy that hasn't traded through multiple market regimes
- Kelly as a ceiling, not a target — using the formula to confirm your position size is nowhere near the point of danger, rather than to set it directly
"Kelly tells you the most you could rationally risk. It was never meant to tell you what you should risk." — the distinction that gets lost whenever the formula gets applied literally.
Why the Formula Is Only as Good as Your Journal
Both Kelly inputs — win rate and average win/loss ratio — come from your trade history, not from a textbook or a backtest that no longer reflects how you actually trade. A journal that logs every closed trade with its outcome is what turns "I think I win around half the time" into a number precise enough to put into a formula. Without that data, running the Kelly calculation is just running real math on a guessed input, which produces a confident-looking number that means nothing.
This is the same reasoning behind calculating position size from your R-multiple history rather than a fixed dollar amount — the formula only helps once the inputs are measured, not assumed.
How Many Trades Before You Can Trust the Inputs
A win rate calculated from eight trades can look like 62% and mean almost nothing — one more win or loss swings it by more than 10 points. Most traders need somewhere around 30 to 50 closed trades on a specific setup before the win rate and average R stabilize enough to be useful in a Kelly calculation, and even then the number should be treated as a moving estimate, not a fixed constant.
- 1Calculate win rate and average win/loss ratio per setup, not blended across your whole strategy mix
- 2Recalculate both inputs every 20–30 trades rather than setting them once and forgetting them
- 3Use half or quarter Kelly as your actual risk-per-trade figure, keeping full Kelly as a theoretical ceiling
- 4Recheck the calculation after any stretch of trades that meaningfully shifts your win rate or R
- 5Treat a negative Kelly output as exactly what it is — a signal the setup currently has no edge to size for
When Kelly Says Zero (or Negative)
If your win rate and R combine to a negative Kelly percentage, the formula isn't glitching — it's telling you the setup, as currently measured, has no statistical edge to size a position around. No amount of clever position sizing fixes a strategy with negative expectancy; the only fix is a better setup, better trade selection, or tighter losses, all of which are exactly what a journal built to expose overtrading and setup drift is designed to surface before the losses compound.
Track every trade with TradeEdge and get your real win rate and win/loss ratio calculated automatically — the two numbers the Kelly Criterion actually needs, without the manual spreadsheet math.