Lesson 46 · The Journal
Four winning trades in a new setup feels like a discovery. It’s mostly noise. The question is how much data turns noise into something you can act on.
Small samples swing a lot. At a 37% win rate, 20 trades can easily show anything from about 15% to 59%.
Treat 20 trades as a hint, 50 as worth testing, 100+ as something to lean on. These are rough guides, not laws.
Trust big differences over small ones, and patterns that repeat across different months.
Every breakdown in this series, by setup, by hour, by price, by mood, has the same trap. When you slice your trades into groups, each group gets smaller, and small groups lie. A setup with five trades and four winners looks brilliant. It could easily be ordinary.
How much can a win rate move by chance?
Here’s the math for a trader whose true win rate is 37%. The range shows roughly where the measured win rate will land most of the time (about 95%), just from luck, depending on how many trades you count:
| Trades counted | Measured win rate will usually land between |
|---|---|
| 10 | about 7% and 67% |
| 20 | about 15% and 59% |
| 50 | about 23% and 51% |
| 100 | about 27% and 47% |
| 140 | about 29% and 45% |
This uses a standard approximation, and real trading is messier, but the shape is what matters: with 10 or 20 trades, almost any win rate is possible by luck alone. The range only tightens as the count grows.
Rough guides for acting on a pattern
| Trades in the group | What it’s good for |
|---|---|
| Under 20 | A hint. Note it, keep going, don’t change anything yet. |
| 20–50 | Enough to design a small test, like a rule you try for the next few weeks. |
| 50–100 | Worth acting on if the difference is large and makes sense. |
| 100+ | Something you can lean on, while still checking it keeps holding. |
These are practical guides, not statistical thresholds. Two things make a small sample more convincing: a very large difference, and a reason that makes sense. “My midday trades are negative and they’re also where my boredom tags are” is more believable than a random-looking gap.
Check it across time
A strong way to test a pattern is to split your history in two, for example the first half and the second half of your trades. If the finding shows up in both halves, it’s more likely real. If it only shows up in one, it might be a single good or bad stretch.
Cutting is safer than adding
Acting on a small sample is less risky in one direction. Stopping a trade type that looks negative, or reducing size there, mostly costs you some opportunity. Increasing size on a setup because of a small hot streak can cost you real money. Be quicker to cut and slower to scale up.
Putting it together
- Small groups swing widely by luck; 20 trades can show almost anything.
- Under 20: note it. 20–50: test it. 50–100: act if the gap is big. 100+: lean on it.
- Big differences with a clear reason are more believable.
- Check a pattern in two separate periods.
- Be quick to cut, slow to size up.
An honest word: no sample size makes the future certain. More trades just make it less likely that you’re fooling yourself with a lucky or unlucky stretch. Educational content only, not financial advice.
Free, by email
Get the journal series in your inbox
All 16 journal lessons in one email, then a short note when a new lesson goes up. Nothing else. Unsubscribe any time.
← Previous: Losing Streaks: What’s Normal and What’s a Warning Next: Turning a Journal Finding Into a Rule You Test →