Concepts5 min read

Win Rate vs Expectancy

Win rate is the percentage of trades that close profitable. Expectancy is the average dollar (or R-multiple) outcome per trade across the full distribution of wins and losses. A strategy can win 80% of the time and still bleed equity if the 20% of losers are large enough to overwhelm the cumulative gains.

The computation

Expectancy combines win rate, loss rate, and the average size of each outcome. The standard formulation:

Expectancy = (W × AvgWin) − (L × AvgLoss)

Where W is win rate (fraction of trades closing positive), L is loss rate (1 − W), AvgWin is the mean profit on winning trades, and AvgLoss is the mean loss on losing trades expressed as a positive number. A normalized form uses R-multiples, where each trade's outcome is divided by the initial risk:

Expectancy_R = (W × AvgWin_R) − (L × AvgLoss_R)

This produces expectancy in units of risk-per-trade, which makes strategies with different position sizes directly comparable.

Interpretation and typical ranges

Positive expectancy is necessary but not sufficient for a viable strategy. A value of +0.10R means each trade returns, on average, 10% of the risk amount. Anything below +0.05R per trade is fragile — small shifts in execution costs, slippage, or regime change push it negative. Robust mean-reversion and breakout systems typically land between +0.15R and +0.40R per trade after costs.

Win rate alone tells you almost nothing. A 90% win rate with an average win of 0.2R and average loss of 3R yields:

(0.90 × 0.2) − (0.10 × 3.0) = 0.18 − 0.30 = −0.12R

The system loses 0.12R per trade despite winning nine times out of ten. Conversely, a 35% win rate with average win of 4R and average loss of 1R yields +0.75R per trade — a strong system that feels terrible to trade because two out of three closes are losses.

The payoff ratio AvgWin/AvgLoss and the win rate are linked through breakeven: a strategy with W win rate breaks even when AvgWin/AvgLoss = (1−W)/W. A 50% win rate needs at least 1:1. A 30% win rate needs at least 2.33:1. A 70% win rate can survive on 0.43:1.

What expectancy does not capture

Expectancy is a first-moment statistic. It says nothing about the variance, skew, or path of returns. Two strategies with identical +0.20R expectancy can have radically different drawdown profiles if one delivers steady small wins and the other concentrates gains in rare outliers.

It also ignores trade frequency. A system with +0.50R expectancy that fires twice a year underperforms a +0.10R system that fires daily, despite the headline number favoring the former. Annualized expected return is closer to Expectancy × TradesPerYear × RiskPerTrade.

Sequence risk is invisible to expectancy. A strategy with positive expectancy can still ruin an account if losing streaks are long enough relative to position sizing. The probability of a losing streak of length k in a strategy with loss rate L is approximately 1 − (1 − L^k)^N over N trades — and this grows quickly for low win-rate systems.

Expectancy computed on in-sample backtest data is almost always inflated by overfitting. Out-of-sample expectancy commonly degrades by 30–60% from the optimized in-sample value. Treat backtested expectancy as an upper bound, not an estimate.

Finally, expectancy assumes independent, identically distributed trades. Real strategies have serial correlation, regime dependence, and execution skew. A trend-following system has clustered wins; a short-volatility system has clustered losses. The arithmetic mean understates tail risk in both cases.

How Kestrel Signal presents it

Kestrel Signal reports expectancy in both dollar and R-multiple form on every backtest summary, alongside win rate, payoff ratio, and trade count. The strategy comparison view ranks systems by R-expectancy rather than win rate by default, and flags any result where the implied breakeven payoff ratio sits within 10% of the observed payoff — the zone where small parameter changes flip the strategy negative. Distribution plots of per-trade R-outcomes are surfaced next to the headline number so you can inspect the full shape, not just the mean.