Win Rate Is Not the Score

Win rate is the share of trades that made money. It is the easiest number to understand and the easiest to misread, because it says nothing about HOW MUCH each win and each loss were. Ten trades that each make $10 and one that loses $200 is a 91% win rate and a $100 loss.

Expectancy is what an average trade earns: (win rate × average win) − (loss rate × average loss). A rule that wins 40% of the time with winners twice the size of its losers earns 0.4 × 2 − 0.6 × 1 = +0.2 of a typical loss on every trade. A rule that wins 70% of the time with winners a third the size of its losers loses: 0.7 × ⅓ − 0.3 × 1 ≈ −0.07. The second one feels better every week and drains the account.

This platform measured what followed every new 20-day high (the NEW_20D_HIGH condition) across 499 S&P 500 stocks. Over the next 20 days the stock did better than the rest of the market just 49.9% of the time, a coin flip. Yet on average it finished 0.31% ahead, because the average winner (+5.99%) ran further than the average loser (−5.35%). Judged by its win rate, the setup looks useless; judged by its expectancy, it is positive. That is why nothing here is shown as a lone 'success rate': even the Chart setups tab's Confidence score only ranks how clean a pattern looks, and the tab says it is not a probability of profit.

The Scanner's Backtest tab shows win rate beside profit factor: everything the winning trades made divided by everything the losing trades lost. Above 1, the rule made money over the test (before costs); below 1, it lost. Read the two together, along with the trade list, and treat any tool, here or anywhere else, that shows a win rate on its own as hiding the other half.

Strategies that win very often usually do it by taking small profits and letting losses run: refusing stops, averaging down, selling insurance. They look superb right up to the one loss bigger than all the wins together. A high win rate is a reason to ask how big the losses get, not a reason to relax.