Why Win Rate Is a Misleading Metric
Success rate, also called win rate, measures how often a trading system produces a profitable trade. It is easy to calculate and psychologically satisfying: most people prefer a system that is right more often than it is wrong. The source notes that evaluating a system mainly by its success rate is a simplistic approach. A system may be correct in 90% of its trades and still lose money when used across a sufficiently large number of trades. The reason is that win rate says nothing about the size of wins relative to losses. If a system wins small amounts frequently but occasionally suffers a much larger loss, the frequent wins can be erased by a single adverse outcome. For example, a generic system that wins +0.1R nine times and loses -2R once has a 90% win rate but a net result of -1.1R. The opposite is also possible: a system with a low win rate can be profitable if its average winner is much larger than its average loser. Trend-following approaches, for instance, often tolerate many small losses while seeking a few large gains. Win rate also ignores the sequence of trades, drawdowns, and the number of opportunities. A 90% win rate over ten trades is not the same as 90% over a thousand trades, and a string of losses at the wrong time can force a trader to abandon the system. Therefore, success rate is a descriptive statistic, not a complete measure of system quality. It can be useful as one input, but it cannot answer whether the system has a positive edge after costs, whether that edge is stable, or how much risk the trader must accept to realize it.
Expectancy: The First Step Beyond Accuracy
Expectancy moves beyond win rate by combining accuracy and payoff into a single average. In R-multiple terms, expectancy is the average result per trade expressed in units of initial risk. A common formula is: Expectancy = (Win rate × Average win in R) − (Loss rate × Average loss in R). If a system wins 40% of the time with an average winner of +3R and loses 60% of the time with an average loser of -1R, its expectancy is (0.40 × 3) − (0.60 × 1) = 1.2 − 0.6 = +0.6R per trade. This is more informative than a 40% win rate alone. However, the source warns that a system with a higher average expectancy is not automatically better. A system averaging +2.3R per trade may look superior to one averaging +0.4R, but the answer is not guaranteed. Expectancy is an average, and averages can hide important details: trade frequency, the distribution of outcomes, the size and duration of drawdowns, and the reliability of the estimate. A high-expectancy system may trade rarely, meaning it produces fewer total R over a given period. It may also have fat-tailed results, where a few large winners dominate the average; remove those outliers and the edge may disappear. Estimation error matters as well. A positive expectancy measured on a small sample may not persist. In addition, expectancy does not indicate how much capital is required to survive the losing sequences that occur within the distribution. So expectancy is a useful building block, but it is not the final scorecard. It answers how much a system earns per trade on average, not how smoothly, how often, or with what risk that average is achieved.
