Win rate vs. expectancy: why the indicators that win most often lose money
In our tests the rules with the highest win rates lost money, and the rules that made money won about a third of their trades. The number that decides the result is not how often you win but how much.
How this was made: the code, the research runs and the first drafts of this text were produced by an AI assistant (Claude) working under the operator's direction; the operator made every decision and checked the numbers. All trading was on paper: real prices, no real money. Nothing here is investment advice.
Two rules, opposite results
On 4-hour bars, the median setting of RSI oversold-buying won 59% of its trades and returned −6.3% a year after costs. The median Bollinger breakout setting won 34% of its trades and returned +52.9% a year. The chart puts every indicator we tested on the same two axes.
Buys strength (trend and breakout) Buys weakness (oversold) Both, depending on the setting
The arithmetic
A strategy's expected result per trade, its expectancy, is:
win rate × average win − loss rate × average loss − costs
A rule can win 63% of the time and still lose if its losses are larger than its wins. Our daily-chart test of 34 textbook rules shows the pattern clearly:
| Rule (daily bars) | Trades | Win rate | Average win | Average loss | Total, two years |
|---|---|---|---|---|---|
| Williams %R, buy above −80, sell above −20 | 314 | 63.1% | +8.3% | −18.4% | −24.9% |
| Stochastic (14, 3), buy above 20, sell above 80 | 228 | 62.7% | +11.7% | −19.0% | −12.8% |
| Bollinger, buy below the lower band | 237 | 61.2% | +9.5% | −14.3% | +1.7% |
| RSI(14), buy when it turns up from 30 | 107 | 57.9% | +10.1% | −17.8% | −6.6% |
| RSI(14), buy above 70, sell below 60 | 118 | 43.2% | +31.5% | −8.1% | +55.3% |
| Keltner, buy above EMA 20 + 2 ATR | 137 | 41.6% | +33.8% | −10.6% | +48.2% |
| Bollinger, buy above the upper band | 237 | 33.8% | +31.1% | −8.8% | +53.0% |
Williams %R won 63.1% of its trades, but its average loss was more than twice its average win, so each trade lost 1.5% on average. The RSI breakout won only 43.2%, but its average win was almost four times its average loss.
Why oversold buying wins often
Small bounces are common, so a rule that buys a dip and sells into the first recovery is right most of the time. The trades that go wrong are the dips that turn into falls, and those are large. In the same daily test, buying when RSI was below 30 was followed by a rise five days later 53.4% of the time, but the average move was only +0.35%, statistically indistinguishable from zero; after 20 days the share of rises, 42.9%, was the same as on any random day.
Why trend rules win rarely
Trend-following rules buy after a rise has begun, cut losers quickly and hold winners for as long as the trend lasts. Most entries are false starts. Our 4-hour breakout rule won 36.4% of 1,037 trades in nine years of replay; the top 10% of trades produced more than twice the total net profit. More on that rule.
How to use this
- Judge a rule by its expectancy after costs, never by its win rate alone.
- Look at the size of the worst losses. A high win rate with small wins often hides a few large losses.
- Expect a good trend rule to lose most of the time, and size positions so you can live through long losing streaks.
Work out the win rate your own numbers need with the break-even win rate calculator.