Profit factor lies on small samples
Thirty trades and a profit factor of 2.1 mean nothing. How many trades the metric actually needs — and what to look at until you have them.
Profit factor is gross profit divided by gross loss. Above one, the system makes money. It reads instantly, which is precisely why people trust it sooner than they should.
Why thirty trades is not enough
Profit factor is a ratio, and both halves behave badly. The numerator holds rare large wins, the denominator rare large losses. On a short history a single trade rewrites the answer.
An example. Thirty trades, gross profit 210,000, gross loss 100,000 — profit factor 2.1, an excellent system. Now remove the single best trade, worth 90,000: 120,000 against 100,000, profit factor 1.2. The method did not change. One trade out of thirty did.
It works the other way too: one large loss that has not happened yet drags 2.1 down towards one. The question is not whether it will happen, but whether it is inside your sample.
How many you need
There is no universal threshold — it depends on how dispersed your results are. Practical guides:
- Under 50 trades — profit factor is an anecdote, not a metric. Look at it, do not lean on it.
- 50–200 — it starts to mean something, provided you have no rare large outliers. If you do, it does not.
- 200+ — trustworthy, but still as an estimate rather than a constant.
A quick robustness check: drop your three best trades and recompute. If profit factor falls below one, you do not have a system, you have a few good days. It takes thirty seconds and is more honest than any confidence interval.
What the metric hides
Order. Profit factor is identical whether losses are scattered or arrive in a row. Ten losses in a row is a drawdown and usually an abandoned system — and profit factor says nothing about it.
Position size. It counts money, not decision quality. One large lucky entry lifts it as much as a hundred careful ones.
Break-even trades. They enter neither half, and quietly improve the picture without earning anything.
What to look at while the sample is small
- The distribution, not the average. A histogram of results says more than any single figure.
- Streaks. The longest run of losses in your history is what you will have to sit through, and it is usually longer than you remember.
- Reshuffles. Run the same trades in random order many times and look at the spread of outcomes. That is the honest answer to "was I lucky or do I have an edge".
- Rule adherence. On small samples, the share of trades taken according to plan tells you more than any return figure.
In MaxProfit the cumulative profit factor is plotted across the history, so you see not the final number but how it moved. A curve that wanders all the way and settles only at the end is a more honest statement about reliability than the figure printed in the corner of a report.
Read next
MFE and MAE: what you left on the table and what you sat through
The result of a trade tells you how it ended. MFE and MAE tell you what happened on the way — and that is usually where the fixable part is.
Expected value per trade: why three calculators give three different answers
The formula is short and everybody knows it, yet everybody gets a different number. Here is where the figures actually diverge — including the six months we showed three of them ourselves.
Maximum drawdown: from the equity peak or from capital?
Two ways of expressing drawdown as a percentage produce numbers that differ several times over. Both appear in reports, and almost nobody states which one they used.