Kelly Criterion Calculator
The maths-optimal fraction of your bankroll to risk per trade, given a known win rate and payoff ratio.
Why half-Kelly is the honest default
The Kelly fraction assumes you know your win rate and payoff exactly. In trading, both are themselves estimates with real uncertainty - a 55% win rate measured on 100 trades could be a true 50% getting lucky, or a true 60% getting unlucky, and the gap matters enormously for sizing.
Full Kelly on an overstated edge means you are betting more than the true Kelly on the true edge - the only region where the long-run growth is negative. Half-Kelly gives up roughly a quarter of the growth and cuts the drawdown risk by much more. For most traders it is the only version worth using.
Common questions
What is the Kelly criterion?
It is the fraction of your bankroll that maximises the long-run growth rate of your capital, given a known win rate and payoff ratio. The formula is f* = (bp - q) / b, where p is your win rate, q is 1 - p, and b is average win divided by average loss. Below the Kelly fraction you grow slower than you could; above it, you grow faster for a while and then blow up with mathematical certainty.
Why do most people use half-Kelly or less?
Because the full Kelly fraction assumes you know your win rate and payoff exactly. In trading, both estimates are themselves noisy — your observed edge over 100 trades could easily be twice or half the true edge. Risking the full Kelly on an overstated edge means risking far more than Kelly on the real edge, which is where ruin lives. Half-Kelly roughly halves the growth rate but cuts the drawdown risk by much more.
What if the Kelly fraction is negative?
A negative Kelly means your average loss exceeds your win rate times your average win — your edge is negative. The optimal bet size is zero. If your own trade history gives a negative number here, no amount of position sizing will save it; the problem is upstream.
Should I use Kelly on my raw backtest numbers?
Not directly. A fresh backtest's win rate and payoff are almost always optimistic — curve-fit edges tend to shrink out of sample. Use Kelly on numbers from a frozen out-of-sample test, or from a long enough forward test that the confidence interval on your average trade excludes zero (that is what the Monte Carlo Simulator and Risk of Ruin Calculator together establish).
Why does Kelly recommend such large bets sometimes?
Because the formula ignores path. A 25% Kelly can be mathematically optimal and still produce a 70% drawdown you cannot stomach. In practice the ceiling that matters is the drawdown you can hold through without closing the account; most working traders cap position size well below Kelly for exactly that reason.
Where we use this
Kelly only works on real numbers. Guessed win rates and curve-fit payoffs will blow up a Kelly-sized account faster than one sized small.
Get the rulebooks: the exact rules and real tested numbers of every strategy we've run. Free by email.