Kelly Criterion — P&L

Optimal fraction of capital to risk per trade. Kelly = W − (1−W)/R, where W = win rate, R = payoff ratio.

Computed from: Trades list
Scope: Single report
Range: 0 – 100%
Direction: Higher is better

Basis

Computed on the Per-Trade P&L Axis (account currency)

Kelly Criterion is the bet size that grows your account fastest in the long run — the fraction of your capital to risk per trade for the quickest possible compounding. It's built from your win rate (how often you win) and payoff ratio (average win ÷ average loss): f* = W − (1 − W) / R (the * just marks it as the optimal fraction). Here's the twist: this mathematically best number is almost always more than you should actually bet — because it assumes your win rate is exactly right and that you can stomach gut-wrenching swings, and neither is true.

How it's calculated

Kelly % = W − (1 − W) / R

W: win rate — the share of trades that are winners
R: payoff ratio — average win ÷ average loss (loss as a positive size)

In this product: Built directly from win-rate and payoff ratio, so it inherits both of their flaws. EquityTruth emits it on the P&L axis (kelly) and the Return axis (kelly_gain), each as a percent. It is not clamped: with no edge it goes negative, which means "don't bet."

On the P&L axis W and R come from wins and losses measured in account currency, so the result reflects how you actually sized trades.

What it tells you

Kelly is a sizing answer, not an edge test. A positive value says "an edge exists, here's the growth-maximizing bet"; the bands below are that bet as a percent of capital — but read every one as "bet a fraction of this" (half-Kelly is the pro default — see below).

Value	Reading	Notes
< 0%	No edge	Win rate and payoff together lose money. Kelly says bet nothing.
0 – 5%	Small edge	A modest edge. Even full Kelly here is already meaningful risk; size below it.
5 – 15%	Solid edge	A real edge — but full Kelly is volatile. Half-Kelly (2.5–7.5%) is the practical range.
> 15%	Large edge	Full Kelly here is reckless — drawdowns are brutal. Quarter-Kelly is plenty.

One subtlety: Kelly gives you a risk budget per trade (how much of capital to put at stake), not a lot size. Your stop distance converts that budget into an actual position — a wider stop means fewer lots for the same Kelly fraction.

Same edge, two bet sizes. Full Kelly (accent) reaches higher but halves the account along the way — from 200 back to 98, a −50% drawdown. Half Kelly (green) ends only a little lower with a fraction of the turbulence. The smoother curve is the one you can actually sit through.

Worked example

A system wins 55% of trades with a payoff ratio of 1.5 (average win 1.5× the average loss):

f* = 0.55 − (1 − 0.55) / 1.5 = 0.55 − 0.30 = 0.25  → "risk 25% per trade"

Full Kelly here says risk 25% of capital on every trade — which is insane in practice: a normal losing streak would carve the account in half and then some. Half-Kelly is 12.5%, still aggressive; quarter-Kelly is 6.25%, which is where many would actually trade it. The formula is correct; 25% is the ceiling, not the plan.

Pitfalls

Pitfalls & caveats

Never bet full Kelly — the headline. Full Kelly maximizes growth but a ~50% drawdown is routine, and over-betting it (more than f*) eventually goes bust no matter how good the edge. Treat the number as a ceiling and bet a fraction of it.
Estimation error overstates it. W and R come from past trades and are noisy; a small or lucky sample produces a Kelly that is far too high. Fractional Kelly is partly a hedge against this — you're betting a fraction of a number you can't fully trust.
Inherits every win-rate and payoff flaw. Outliers in the average win or loss move R; a thin sample moves W. It also assumes trades are independent, which forex rarely honors (correlated pairs and grids are one bet wearing many tickets).
Assumes the edge persists. Kelly sizes for the edge you had. If the edge decays, you're over-betting a strategy that no longer pays.
Most retail traders are already past it. Sizing by feel under 1:100–1:500 leverage, the typical retail forex account is already over-betting relative to Kelly, often several times over. For most people the lesson isn't "reach Kelly" — it's "you're already well beyond it."
It's sizing, not edge-detection. A negative Kelly means there's no edge — period. To judge whether the edge is real, use expectancy and SQN, not Kelly.

Fractional Kelly

Almost no one bets full Kelly. Bet half of Kelly and you still capture about 75% of the growth — with far gentler swings (the growth math: betting a fraction k of full Kelly gives k(2 − k) of the maximum, so k = 0.5 → 0.75). The curve gets dramatically smoother for only a quarter less growth.

The same math explains why you should err low when unsure. Push the bet the other way — double Kelly (k = 2) gives 2 × 0 = 0 growth: you hand your entire edge back to volatility, and anything past that compounds negatively despite a real edge. Under-betting costs you a little growth; over-betting can cost you everything. Since your win rate and payoff are only estimates — and could be too rosy — the safe direction to be wrong is down. That's the real reason pros size between quarter-Kelly and half-Kelly, never full.

Win Rate and Payoff Ratio are the two inputs Kelly is built from, Expectancy tells you whether the edge exists (Kelly assumes it does), and Max drawdown is the pain full Kelly inflicts — the reason to bet a fraction.

Built fromWin Rate Payoff Ratio