How Many Trades Before It Means Anything?
Ten trades can make a coin look like a strategy. How far a win rate wobbles from luck alone, and why bots here wait for 30 trades.
Reading a BacktestIntermediate 6 min read· Lesson 3 of 6 in Reading Results Honestly
A coin flip with a track record
Flip a fair coin ten times and it lands heads seven or more times about 17% of the time, roughly one run in six. Call heads a 'win' and you have a 70% win rate from something with no edge at all. Small samples don't just add noise; they produce convincing stories.
How wide luck runs
For a rule whose true win rate is 50%, here is the range luck alone produces 95% of the time:
| closed trades | win rate you might see |
|---|---|
| 10 | 20% to 80% |
| 20 | 30% to 70% |
| 30 | 33% to 67% |
| 100 | 40% to 60% |
| 400 | 45% to 55% |
Telling a 55% rule from a 50% one with any confidence takes hundreds of trades, not dozens.
Count independent bets, not rows
Ten trades that all entered in the same week are closer to one trade than to ten: they rode the same market. So are twenty stocks a rule bought on the day the whole market jumped. The independent evidence is usually smaller than the trade count says.
On this platform
The Trading bots page ranks no bot until it has closed 30 trades, and until then the bot's page says plainly that whatever its return shows is mostly luck, good or bad. A backtest's trade count is in its summary: a rule that fired four times on one stock has told you about four days, not about the rule.
What to do with a small sample
Treat it as a reason to look further, never as a result. Test the same rule on more stocks and more years, paper trade it, and keep real money out until the range in the table above is narrower than the edge you think you have.
Where this lives in ScanPro
Stock pages are open to everyone; the rest need a free account.