Win Rate Significance Calculator
Enter your wins and total trades to see whether your win rate is actually distinguishable from chance — or from any baseline you choose — using a Wilson score confidence interval and an exact binomial test. The question this site exists to ask about every pattern, applied to your own trade record.
Trade record
50% for "is this better than a coin flip?" Use your breakeven win rate (from the risk/reward calculator) to test "is this beating what I need to be profitable?"
Method
Wilson score interval (Wilson 1927)
Exact two-sided binomial test
Significance
observed win rate over 100 trades
- 95% confidence interval
- 48.2% – 67.2%Wilson score interval
- Exact two-sided p-value
- 0.1332vs. 50% baseline
Blue band = confidence interval · white line = your observed rate · green line = baseline
A "not distinguishable" result does not mean your win rate is 50% — it means this sample size can't rule out chance producing a result at least this extreme. More trades narrow the interval; the interval width shrinks roughly with the square root of the sample size, not linearly.
Why win rates need error bars
A strategy that wins 58% of 100 trades sounds like a real edge. Run the exact binomial test against a 50% baseline and the two-sided p-value comes out to about 0.13 — meaning a coin-flip strategy would produce a result at least this extreme roughly one time in eight, just from sampling noise. That's not proof of no edge, but it's nowhere near strong enough evidence to call it skill rather than luck.
The same 58% win rate over 1,000 trades tells a very different story: the 95% confidence interval narrows from roughly 48%–67% (at n=100) to about 55%–61% (at n=1,000), and now excludes 50% entirely. Same win rate, opposite conclusion — because sample size, not the headline percentage, is what determines whether a win rate means anything.
This is the same statistical discipline the site applies to seasonal patterns and congressional trades: a win rate without a sample size and a confidence interval is not evidence, it's a headline.
Frequently asked questions
Is a 60% win rate over 20 trades statistically significant?+
No. At 12 wins out of 20 trades, the 95% Wilson confidence interval runs from roughly 38% to 78% — comfortably including 50%, so a fair coin could easily have produced this result. The exact two-sided binomial p-value against a 50% baseline is about 0.50 (far above the conventional 0.05 threshold). Small samples produce wide intervals; a 60% win rate needs several hundred trades before it reliably excludes 50%.
Why use the Wilson score interval instead of the normal approximation?+
The standard "Wald" interval (p̂ ± z·√(p̂(1−p̂)/n)) has poor coverage for small samples and for win rates near 0% or 100% — it can produce nonsensical bounds like a negative lower limit. The Wilson score interval (Wilson, 1927) corrects for this and is the interval statisticians generally recommend for binomial proportions at any sample size (Brown, Cai & DasGupta, Statistical Science, 2001). This calculator always uses Wilson, never Wald.
What does the p-value actually mean here?+
It is the probability of seeing a result at least as extreme as your observed win rate, if the baseline rate you tested against were actually true. A p-value of 0.13 means: if your true win rate really were the baseline, you'd see a result this extreme (or more extreme) about 13% of the time by chance alone — not rare enough to conclude your edge is real rather than noise, by the conventional 5% threshold.
What baseline should I test against?+
50% tests whether you're better than a coin flip, which is a low bar for a directional strategy. A more useful test is your breakeven win rate — the win rate your risk/reward ratio requires just to break even, which the risk/reward calculator computes for you. Testing against your breakeven rate answers the real question: is my edge distinguishable from exactly breaking even?
How many trades do I need before a win rate becomes meaningful?+
It depends on how far your observed rate is from the baseline — a large edge shows up with fewer trades than a small one — but confidence intervals narrow with the square root of sample size, not linearly, so going from 25 to 100 trades only halves the interval width, not quarters it. As a reference point, distinguishing a 55% win rate from 50% at 95% confidence typically needs several hundred trades; distinguishing 70% from 50% needs far fewer, as the "70/100 vs. 50%" example on this page shows.
Related tools
Risk/Reward Ratio Calculator
Find your breakeven win rate — a better baseline to test against than 50%.
Kelly Criterion Calculator
Turn a genuinely significant win rate into an edge-optimal stake.
Backtesting Engine
Get a real win/loss count from historical data instead of guessing.
All Trading Tools
Browse every free tool on QuantEngines.
This calculator performs a Wilson score interval and an exact binomial test on the win/loss counts you provide. It is not investment advice, does not evaluate any specific strategy or trade, and a statistically significant win rate is not a guarantee of future performance.