62.5%
Example: 25 wins from 40 trades
62.5%
Example: 25 wins from 40 trades
47.0% to 75.8%
the real win rate, at 95% confidence
Probably above breakeven
The breakeven point for your model is at a 50% win rate with a 1:1RR ratio, which is in the lower part of your range. About 94.3% of the time your model will exceed break even, which is very good.
You've backtested your trading model a number of times and know exactly what your win rate is. This tool shows you how far that number could be from the truth.
Let's say you backtested 40 trades and 25 of them were winners. That is a 62.5% win rate. However, with only 40 backtested trades, your strategy's real win rate could fall anywhere in the range from 47.0% to 75.8% and still be within tolerance of a normal result.
Yes, shocking... right? The real answer to "what is my win rate?" should be its expected range and not just a single number. The shaded bar in our chart is exactly that range.
Your edge is whether your strategy makes money over many trades, and win rate is only half of it. The other half is how much you make on a win compared with how much you lose on a loss. A 40% win rate with a 1 to 3 risk to reward beats a 70% win rate at 1 to 0.33. The breakeven line is there so you can check your real win rate range against the one your risk to reward to needs. The guide on how many trades to backtest for reliable results works through breakeven and costs in full.
The range used in this tool is derived from the Wilson score interval for a proportion. With p as your observed win rate, n as your total number of backtested trades and z as the value for your confidence level:
( p + z²/2n ± z · √( p(1−p)/n + z²/4n² ) ) / ( 1 + z²/n )
Breakeven is 1 / (1 + R), where R is your reward to risk ratio.
This tool runs backward: it starts from a result that you measured and asks which true win rates could have produced it. The simple normal approximation, called Wald, is the wrong tool for that question. It under-covers at small samples, it can spill past 0% and 100%, and it collapses to a zero width range if you won or lost every single trade. Wilson is asymmetric, stays inside 0 to 100, and holds close to its stated confidence down to about 15 trades.
The sample size tool runs in the other direction. Meaning it goes forward from an assumed win rate, and Wald is the better choice there.
"Inside the range" would be too vague, because breakeven just inside of a range and breakeven in the centre of the range are very different results. So the labels comes from how often a strategy that only broke even would have produced a result like yours. It is measured in standard errors, the same unit the range is built from, so the label always agrees with the bar.