How many trades before a track record means anything?
2026-10-12
Most people evaluate a strategy too early. A series of profitable results can feel like proof, but it usually is not. This note explains how much data is needed before a result is meaningful and how we address that limitation in our own live record.
Why a small sample lies
That sounds like a statistical edge. Over a small number of trades, however, luck can produce a 60% rate as easily as skill. The smaller the sample, the wider the plausible range for the true rate.
The arithmetic
For a win rate, the 95% margin of error is approximately 1.96 ? sqrt(p(1 ? p) / N), where p is the win rate and N is the number of trades. The worst case (a win rate near 50%) simplifies to about 0.98 / sqrt(N).
| Trades | 95% margin |
|---|---|
| 25 | ± 20% |
| 100 | ± 10% |
| 400 | ± 5% |
| 1,000 | ± 3% |
A 60% win rate over 25 trades therefore means the true rate could plausibly be anywhere from about 40% to 80%. Over 400 trades, the range narrows to roughly 55%–65%. That is the difference between a guess and an estimate.
What a short record can tell you
- Whether costs are included, or the numbers are gross.
- Whether losing periods and falls from a peak are shown, not buried.
- Whether the method is stated before the results, rather than reverse-engineered after them.
- Whether you can check that the record was not quietly edited.
What it cannot
Whether the edge is real. That requires time and a sufficient number of trades; there is no shortcut, and claims to the contrary warrant skepticism.
How we handle it
Our live record is still short. We publish the trade count, the return, maximum drawdown, and a clear statement that it is too early to draw firm conclusions. The sample can be followed on the performance page rather than judged from a single image.
This is research and education, not investment advice. Trading cryptocurrency involves substantial risk of loss.
See the published record
The public record is free and read-only: daily close results, drawdowns, and methodology, with no sign-in required.