A win rate is the share of trades that closed green. It is the number every strategy seller leads with, and it is close to worthless on its own. Not misleading in a subtle way — worthless in a way you can measure. Here is the measurement, run on our own data.
The test
In August 2026 we swept 9,363 strategy configurations against real S&P 500 futures history: nine different entry styles, four timeframes, a range of stops and profit targets, every one of them keeping only the results with at least 15 closed trades. For each configuration we recorded two numbers — its win rate, and its profit factor.
Profit factor is gross winnings divided by gross losses. Above 1.0, the winners paid for the losers and the system made money. Below 1.0 it lost money. It is the plainest possible measure of “did this work.”
Then we asked the only question that matters: if you know a system's win rate, how much do you know about whether it made money?
The answer: almost nothing
The correlation between win rate and profit factor across all 9,363 configurations was 0.12. Squared, that means win rate explained 1.5% of the variation in whether a system made money. The other 98.5% is somewhere else entirely.
9,363 configurations · hypothetical engine results on ES futures data
Dashed line = the average across all bands (8.7%). Amber = the bands people advertise.
Read the shape, not the bars. It is flat. A system that won 80% of its trades was profitable 16.5% of the time. A system that won barely half was profitable 11.4% of the time. Going from a coin-flip win rate to an outstanding one bought you about five percentage points of extra chance that the thing made any money at all.
The median profit factor tells the same story even more bluntly. It sits between 0.51 and 0.64 in every single band. Whatever the win rate, the typical configuration lost money at about the same rate.
| Win-rate band | Configs | Made money | Median PF | PF range |
|---|---|---|---|---|
| under 40% | 2,789 | 3.2% | 0.58 | 0.01–1.50 |
| 40–49% | 2,211 | 10.9% | 0.64 | 0.03–1.97 |
| 50–59% | 1,992 | 11.4% | 0.59 | 0.04–2.24 |
| 60–64% | 791 | 10.5% | 0.51 | 0.05–2.48 |
| 65–69% | 590 | 8.6% | 0.57 | 0.09–1.97 |
| 70–79% | 662 | 10.4% | 0.57 | 0.05–2.71 |
| 80% and up | 328 | 16.5% | 0.64 | 0.20–2.77 |
Why this happens
Win rate is not a measure of quality. It is a design choice, and you can set it almost anywhere you like before you have tested anything.
Every trade has a stop (where you get out for a loss) and a target (where you get out for a gain). Move the target closer and you win more often, because you are asking the market for a smaller favour. Move it further and you win less often but earn more each time. The win rate is mostly just a readout of where you put the target.
So a 75% win rate can mean either “this system is excellent” or “this system takes tiny profits and occasionally eats an enormous loss.” The number itself cannot tell you which, and the second kind is far more common — of the 1,580 configurations in our sweep that hit a 65% win rate or better, 89% lost money.
Their profit factors ranged from 0.05 to 2.77. Same headline win rate. Completely different outcomes.
What to ask instead
When anyone shows you a track record — including ours — these five questions do the work that a win rate pretends to:
- What is the profit factor? Gross wins over gross losses. Under 1.0 the system lost money no matter how good the win rate sounds.
- What was the worst drawdown? The deepest peak-to-trough fall, in dollars. This is the number that decides whether you can actually sit through the system, and it is the one most often left out.
- How many trades? Under about 30, a win rate is noise. Ten trades can look spectacular by luck alone.
- Was it tested on data it wasn't built on? Any strategy can be tuned to look perfect on the stretch of history used to build it. The only meaningful test is a period the designer did not get to look at.
- How many versions were tried before this one? This is the question nobody asks and it may be the most important. Test enough variations and something will look brilliant by pure chance — we found 74 configurations that passed a serious robustness screen, which is roughly the number random noise produces at that search width. The 75th finding is not a discovery.
We had to learn this on ourselves
This is not a lesson we read somewhere. In August 2026 the system running as our flagship looked good on the timeframe it had been selected on: 57% of trades green, profit factor 2.93. Tested against a full year of hourly data it won 20% of its trades with a profit factor of 0.48, and only 3 of 12 months finished green.
It had been fitted to its own test window and we had not checked hard enough. We retired it rather than re-tuning it, because re-tuning a curve fit just produces a better-hidden curve fit. The replacement was chosen by a different rule: not “which configuration scores best” but “which configuration still works when you nudge every setting,” which is much harder to fake.
You can read what the current system was validated at, including the numbers that are unflattering, on the demo desk. When it has a live record worth showing, that goes up too — including the losing months.
The short version
A win rate on its own tells you where someone put their profit target. It does not tell you whether they made money. Ask for the profit factor, the drawdown, the number of trades, and what happened on data the system was not built on — and be suspicious of anyone who leads with the win rate and goes quiet on the rest.
Every figure on this page is hypothetical output from our own backtest engine over historical ES futures data. It is not a track record, not a prediction, and not a promise of results. New to the vocabulary? The glossary has plain definitions.