Bet too small and your edge barely compounds. Bet too big and a normal losing streak wipes you out. The Kelly criterion is the math that finds the line between the two — and then, if you’re smart, you stand well back from it.
The Kelly formula for trading
Kelly answers one question: what fraction of capital, risked per trade, maximizes long-run growth? For a system that wins a multiple R of what it risks and loses 1R when it fails, the Kelly fraction is:
f = W − (1 − W) / R
where W is your win rate and R is your average winner divided by your average loser, both in R-multiples. Read it as edge over odds: the numerator is your expected profit per unit risked; you divide by the payoff to convert it into a fraction of the account.
A system that wins half its trades at 2R gives f = 0.5 − 0.5/2 = 0.25 — Kelly says risk a quarter of the account per trade. That number alone should make you flinch, and it should. Pull your real W and R from actual trades using the expectancy calculator; if R-multiples are unfamiliar, the R-multiple explainer covers the foundation.
What full Kelly actually optimizes
Full Kelly maximizes the expected logarithm of wealth — the geometric growth rate. Over an infinite series of bets with known, stable probabilities it compounds capital faster than any other fixed fraction, and it never risks total ruin, because you always bet a fraction and never the whole stack. Those are genuine, provable properties.
The catch lives entirely in the assumptions: known probabilities, a stable edge, infinitely divisible position sizes, and an infinite horizon. A prop challenge violates every one of them. Kelly is the right ceiling to think about, not the number to type into your order ticket.
Why full Kelly’s drawdowns are brutal
Growth-optimal is not comfort-optimal. At full Kelly you should expect to routinely give back a large share of the account — deep drawdowns are a feature of the growth-optimal path, not a malfunction. Two facts make this worse for a real trader:
- You feel drawdowns; log-utility doesn’t. Kelly is mathematically indifferent between a 50% drawdown and its eventual recovery. Your psychology, and your prop firm’s rulebook, are not.
- Over-betting is asymmetric. Bet at twice the Kelly fraction and your long-run growth rate collapses toward zero while your variance explodes — all the pain of aggression, none of the growth. Under-betting merely slows you down.
That asymmetry is the entire reason experienced practitioners size below Kelly on purpose.
Half- and quarter-Kelly in practice
The standard fix is fractional Kelly: multiply f by a constant well under 1. Because the growth curve is flat near its peak while variance keeps falling as you shrink the fraction, backing off costs you little growth and buys a lot of smoothness.
| Fraction of Kelly | Growth captured | Drawdown character |
|---|---|---|
| Full (1×) | Maximum | Deep and frequent |
| Half | Most of the maximum | Roughly halved |
| Quarter | Noticeably lower | Much smoother |
Half-Kelly is the common default for traders who still want to compound; quarter-Kelly suits anyone who values survival over speed — which, on a funded account, is nearly everyone. Convert whatever fraction you settle on into an actual contract or lot size with the position size calculator.
Kelly with uncertain, drifting inputs
Here’s the part the textbook skips: Kelly assumes you know W and R. You don’t. You estimate them from a finite, noisy sample, and they drift as market regimes change. Plug an over-optimistic win rate from twenty trades straight into the formula and you get an f that is not just wrong but dangerously high — because Kelly is most sensitive exactly where your estimate is least reliable.
The practical response:
- Estimate W and R from as many real trades as you have, and treat each as a range, not a point.
- Feed the conservative end of that range into the formula — size on the pessimistic edge, not the hopeful one.
- Understand that fractional Kelly is partly a hedge against estimation error, not only against variance.
This is where a live edge measurement with a Wilson confidence interval earns its keep: instead of “52% win rate,” you get “52%, and the true rate is very likely between 44% and 60%.” The lower bound of that interval is the number Kelly should actually see.
Prop-firm limits as a hard cap over Kelly
On an evaluation or funded account, Kelly is almost never your binding constraint — the firm’s drawdown is. Even a modest quarter-Kelly fraction can exceed what a prop drawdown tolerates across a bad streak, and the firm does not care that your sizing was theoretically growth-optimal. A breach is a breach. Run the two calculations in order:
- Compute a fractional-Kelly size as your growth ceiling.
- Compute the size that keeps a plausible losing streak inside the firm’s drawdown floor — check it with the prop-firm drawdown calculator.
- Take the smaller. Always.
Confirm the exact drawdown mechanics with your firm first — trailing versus static changes step two considerably. This is where hard risk limits enforced at the broker turn theory into a guardrail: set your own per-trade and daily cap at the smaller of the two numbers, and it trips before the firm’s line ever does. Auto-journaling keeps the W and R feeding your Kelly math honest, and if you run the same edge across several accounts, copying keeps the sizing identical on all of them.
Related: Position size calculator · Expectancy calculator · Prop-firm drawdown calculator