What is Sample Size Rule?
Reviewed by Karthikeyan Anbarasan (Founder, TradeLyser) · Updated 19 Oct 2026 · Editorial policy
Why small samples mislead
Each trade outcome is partly luck. Over a few trades, luck can dominate: a setup with a true 50% win rate will often show 30% or 70% over ten trades. As trades accumulate, the luck averages out and the measured figure settles towards the true one. Figure 1 shows how fast that happens. The margin halves roughly every time the number of trades quadruples, which is why the jump from 10 to 30 trades matters so much, and why going from 100 to 300 adds less. Losing streaks are part of the same arithmetic. A setup that wins half its trades will, over a few hundred trades, very likely produce a run of five or six losses in a row at some point, and that run says nothing about whether the edge has gone. Judging a setup on its last streak, good or bad, is judging it on a sample of five.
How many trades is enough
Worked example: is a 60% setup real?
Illustrative. A setup on Nifty futures targets 1R and risks 1R, so it needs a win rate above 50% to make money before charges. It is winning 60% of the time. Whether that is good enough depends on how many trades are behind it. At 30 trades the plausible range still includes 50%, so the setup could be breakeven and have been lucky. Only at around 100 trades does the range clear 50%, and then only just; at 300 trades it clears it with room to spare. Charges push the breakeven a little above 50%, which makes the small samples even less conclusive.
| Trades | Measured win rate | 95% range | Clearly above 50%? |
|---|---|---|---|
| 10 | 60% | About 30% to 90% | No |
| 30 | 60% | 42.5% to 77.5% | No |
| 100 | 60% | 50.4% to 69.6% | Only just |
| 300 | 60% | 54.5% to 65.5% | Yes |
Expectancy needs even more trades
A measured expectancy of +0.3R over 30 trades could really be anywhere from −0.24R to +0.84R, so even the sign is uncertain. At 100 trades the range is +0.01R to +0.59R, which finally excludes a losing setup. Strategies with occasional large winners have a larger spread of results and need more trades still. These ranges assume trades are independent and the market has not changed character; when either is false, the real uncertainty is wider. The practical point is that win rate settles faster than expectancy. A trader can be fairly sure after 100 trades that a setup wins more often than it loses and still be unsure how much it makes per trade, which is the number that decides position size. That is why size increases should wait longer than keep or pause decisions.
Using the rule in your journal
- Tag every trade with exactly one setup, so each setup has its own count.
- Show the trade count next to every win rate, expectancy and profit factor you read.
- Make keep, resize or pause decisions in the 45-minute monthly review, using each setup’s last 30 or more trades.
- Start a new count when you change a setup’s rules; the old trades no longer describe it.
- Keep rule-break trades tagged, so you can see whether poor numbers come from the setup or from execution.
Reference guide
| Context | Value | Reading |
|---|---|---|
| Setup judged after about 30 tagged trades; size raised only after about 100 | Patient | Direction first, then confidence, before more money depends on it |
| Setup dropped after five losses in a row | Too early | Streaks of five or more are normal even in setups with a real edge |
| 200 trades in total, but only 8 on the setup being judged | Wrong count | The rule applies per setup tag, not to the whole journal |
Questions
How many trades do I need to evaluate a trading strategy?
About 30 per setup to read the direction of the results, and around 100 to judge the size of the edge with reasonable confidence. Strategies with rare large winners need more.
Is 30 trades enough to trust my win rate?
Only roughly. At 30 trades the 95% margin around a 50% win rate is about ±18 points, so a 50% measured win rate could really be anywhere from about 32% to 68%.
How is the margin of error on a win rate calculated?
Approximately 1.96 × √(p × (1 − p) ÷ n), where p is the win rate and n the number of trades. At 100 trades and 50%, that is 1.96 × 0.05 = ±10 points.
Does the sample size rule apply to each setup separately?
Yes. If a journal has 200 trades but only 8 on one setup, the numbers for that setup are still based on 8 trades.
Should I stop trading a setup after a losing streak?
Not on the streak alone. Check whether the trades followed the rules and how many trades the setup has in total; decide in the monthly review once it has 30 or more.