Win rate is the fraction of closed trades that end with a profit.
Win rate is the fraction of closed trades that end with a profit. It is a simple descriptive statistic, but on its own it does not measure whether a strategy is economically attractive or statistically reliable.
The main limitation is that win rate counts outcomes without considering their size. A strategy can post many small profitable trades and still lose money overall if its losing trades are larger than its winners. The reverse is also possible. A strategy can have a low win rate and still be profitable if its average winning trade is large enough relative to its average losing trade. For that reason, win rate says little about expected trade value unless it is interpreted together with payoff magnitude.
A basic way to see the mechanism is to combine win probability with average win size and average loss size. Expected value per trade depends on all three. Holding only win rate fixed leaves the economic result undetermined. Two strategies with the same fraction of winning trades can have very different average trade P and L, cumulative P and L, and drawdown behavior because their gain and loss distributions differ.
Metrics that incorporate magnitude are more informative. Profit factor compares gross profits with gross losses, so it reflects whether winning trades outweigh losing trades in aggregate rather than just in count. Average trade P and L measures the mean outcome per trade, which links directly to the strategy's expectancy. These measures answer questions that win rate cannot answer by itself.
Risk also matters. A strategy with an appealing win rate can still have unstable returns, large drawdowns, or weak evidence after accounting for multiple testing. Sonar's glossary defines the deflated Sharpe ratio as a version of the Sharpe ratio adjusted for non normal returns and for the number of trials, with the goal of reducing false discoveries when many variants are tested. This matters because a high win rate does not show whether apparent performance is statistically robust. A strategy can win often in sample and still fail to clear a more demanding risk adjusted standard.
The same issue appears in backtest overfitting. Sonar's backtest overfitting audit is built to test whether backtest results are likely to survive out of sample rather than merely fit historical noise. That framing is important because win rate is an in sample summary of realized trade outcomes. It does not tell you whether the underlying edge is persistent or whether the strategy was selected from many alternatives that looked good by chance. A strategy with a high win rate can still be overfit, and a strategy with a lower win rate can be more credible if its out of sample behavior and adjusted statistical evidence are stronger.
For evaluation, win rate is best treated as a secondary descriptor of trade distribution, not as a standalone score. To judge a strategy, pair it with measures of payoff size, aggregate profitability, and risk adjusted validity. At a minimum that means looking at average trade P and L, profit factor, and a statistic such as the deflated Sharpe ratio, then checking whether the result survives a backtest overfitting audit. Only that combination begins to answer the question that win rate alone cannot: whether the strategy's apparent edge is both economically meaningful and statistically believable.
Covered in depth in the Strategy research fundamentals pillar hub.