The question the test answers
A strategy has an edge if its average trade makes money after costs. For a given payoff there is a win rate at which the average trade is exactly zero, the break-even win rate: 40% when winners are 1.5 times the losers, 50% at equal size, and higher once costs are included. A strategy with no edge wins about that share of its trades, but not exactly, because wins and losses arrive at random. The test asks how often a no-edge strategy would win at least as many of your trades as you did. That chance is the p-value.
With 48 wins in 100 trades at 1.5× winners, the p-value is 6.4%: about one no-edge strategy in 16 would do that well. That is suggestive, not conclusive, and the page says which word fits your number. The p-value is not the chance that you have an edge, and it does not say how big the edge is: for that, read the range.
The exact test, not an approximation
The wins of a strategy whose win rate is the break-even rate follow a binomial distribution, and the p-value is the sum of its upper tail: the chance of 48 or more wins in 100 trades at 40%. The chart on the page draws that distribution and marks your result. No normal approximation is used, so it stays right for short records and for win rates far from one half, where the shortcut fails. The tests behind this page check the tail against a second method and against 200,000 simulated no-edge records.
How much luck can explain: p-values by record length
The same win rate means very different things at different record lengths. Break-even is 40% (winners 1.5 times the losers); the table shows how often luck alone produces each result:
| Observed win rate | 20 trades | 100 trades | 300 trades |
|---|---|---|---|
| 45% | 40.4% (9 wins) | 17.9% (45 wins) | 4.4% (135 wins) |
| 50% | 24.5% (10 wins) | 2.7% (50 wins) | under 0.1% (150 wins) |
| 55% | 12.8% (11 wins) | 0.2% (55 wins) | under 0.1% (165 wins) |
A 50% win rate on 20 trades is entirely ordinary for a strategy with no edge (24.5%), the same rate on 100 trades is unusual (2.7%) and on 300 trades it is very unlikely (under 0.1%). More trades, not a better month, is what turns a hope into evidence.
How many trades do you need?
Suppose your win rate stays at its observed value. The number of trades needed before the whole 95% range for your true win rate sits above break-even is:
| Observed win rate | Trades before the 95% range clears the 40% break-even |
|---|---|
| 45% | 369 |
| 50% | 93 |
| 55% | 41 |
| 60% | 24 |
The edge you are trying to prove sets the work: a 50% win rate needs 93 trades, while 45% needs 369. A thin edge is hard to tell from luck without a long record, which is also why a thin edge is hard to trade with confidence. Keep every trade in a trading journal to get there, and see how to keep a trading log.
Costs raise the bar
Commission and spread make every win smaller and every loss larger, so the break-even win rate rises and the same record is less impressive. For 50 wins in 100 trades with winners 1.5 times the losers:
| Costs per trade | Break-even win rate | Chance luck alone does this well (50 wins in 100) | Expectancy at 50% |
|---|---|---|---|
| 0R | 40% | 2.7% | 0.25R |
| 0.1R | 44% | 13.4% | 0.15R |
| 0.2R | 48% | 38.2% | 0.05R |
Add your costs per trade, in the same R units, in the optional field. The risk reward calculator explains the break-even arithmetic.
The trap of many strategies
A p-value below 5% means luck would do this well one time in twenty. If you test many strategies, settings or markets and keep the best, you will find one by luck. The chance that at least one of independent no-edge strategies looks significant at the 5% level:
| Strategies tried | Chance that at least one looks significant (below 5%) by luck alone |
|---|---|
| 1 | 5.0% |
| 5 | 22.6% |
| 10 | 40.1% |
| 20 | 64.2% |
| 50 | 92.3% |
So a good result found after trying 20 variations needs a much smaller p-value, or a fresh set of trades the strategy has not seen, before you trust it. The same applies to stopping the test when the result looks good: decide the number of trades in advance.
What the verdict means
| Verdict | p-value | What to do with it |
|---|---|---|
| Strong evidence | below 1% | Luck alone would rarely do this. Check the range for how big the edge could be, and that the trades were independent and the rules fixed. |
| Suggestive | 1% to below 10% | Worth continuing, not worth sizing up. Collect the trades the page says you need. |
| Not enough evidence | 10% or more | Luck explains the record easily. Keep trading small and keep records. |
| No edge shown | win rate at or below break-even | The record is no better than no edge, or worse. Fix the strategy or the costs before anything else. |
What the test does not know
- It treats your average winner and loser as exact and the trades as independent. Real payoffs vary, and trades cluster, so real uncertainty is larger than shown.
- It tests the past. A strategy can have had an edge and lose it, and an edge in one market need not carry to another.
- It does not see how the results were chosen. A record that leaves out bad periods or trades cannot be tested honestly.
- It is not advice. Use it to decide how much to trust a record, and size positions with the position size calculator and the risk profile quiz for the risk that fits.
The formulas
p-value = P(X ≥ wins), X ~ Binomial(n, q)
95% range for the win rate (Wilson) = (p + z²/2n ± z × √(p(1 − p)/n + z²/4n²)) ÷ (1 + z²/n), z = 1.96
expectancy at win rate p = p × W − (1 − p) × L − c
trades needed = the smallest n′ for which the Wilson lower bound at your observed rate exceeds qW and L are the average win and loss in R, c the cost per trade in R, p the observed win rate.
Worked examples
48 wins in 100 trades, winners 1.5×
Break-even is 40%. Luck alone gives 48 or more wins about 6.4% of the time. The 95% range for the win rate is 38% to 58% and the expectancy range is −0.04R to +0.44R, which includes zero.
The same win rate on 300 trades
135 wins in 300 is the same 45% as 45 in 100, but the chance of luck alone doing this well falls from 17.9% to 4.4%.
The same record with costs
50 wins in 100 at 1.5× winners. With 0.2R costs the break-even rate rises to 48% and the chance luck does this well rises from 2.7% to 38.2%.
Frequently asked questions
How do I know if my trading edge is real?
Compare your record with what luck alone would produce. If a strategy with no edge would rarely win as many of your trades as you did (a small p-value), and the 95% range for your win rate sits above break-even, the edge is more likely real. A short record almost never settles it.
What is a p-value in trading?
The chance that a strategy with no edge, one whose true win rate is the break-even rate, would win as many of your trades as you did or more. It is not the chance that your edge is real.
What is the break-even win rate?
The win rate at which the average trade is exactly zero for your payoff and costs: (loss + costs) ÷ (win + loss). At 1.5× winners it is 40%, at equal size 50%.
How many trades do I need to prove an edge?
It depends on the size of the edge: the page shows the number at which the 95% range for your win rate clears break-even if your observed win rate holds. A 50% win rate at 1.5× winners needs about 93 trades; 45% needs about 369.
Why does the test treat my average win as exact?
Because the average payoff is a summary of the record and the test is about the win rate. If your wins and losses vary a lot, the real uncertainty is larger, which makes the verdict a best case.
Is this financial advice?
No. It is a statistical description of the numbers you enter. Past results do not predict future results, and trading carries a risk of loss.
An edge has to be found before it can be tested.
These indicators help you define rules you can then test on a record of trades.
Last updated November 2026. For education only. Not financial advice. Results depend on your inputs and cannot predict future performance.