Chart Reporter

How much confidence should I place in a small trade sample?

By Chart Reporter Reviewed 2 min read

The short answer

A small sample can describe what happened without establishing a stable trading tendency. Count eligible observations, inspect concentration and preserve how the sample was selected. Repeatedly changing filters until a strong-looking result appears can make an accidental pattern look convincing.

Look beyond the percentage

A displayed 75% can mean three successes from four observations or 300 from 400. Those samples provide very different amounts of information even though the headline percentage matches. The time period and dependence between observations matter as well as the count.

Trading records are not automatically independent experiments. Several positions can share the same instrument, session or market event. Treating every related outcome as wholly new evidence can exaggerate confidence. A journal makes those groupings visible, but it does not remove the underlying dependence.

Inspect what the calculation omitted

Some analyses require initial stops, excursion data or sufficient candle history. Their eligible count can be smaller than the total trade count. A clean-looking result may therefore describe a subset with unusually complete capture rather than the intended strategy as a whole.

For example, if recent trades have replay context and older trades do not, an excursion analysis may mainly reflect the recent environment. Record that coverage difference before comparing periods. Missing data should lead to a narrower claim, not an assumption that excluded trades behaved like included ones.

Write the question before filtering

Choose a specific comparison and preserve its filters before examining the result. If you discover a promising subgroup during exploration, label it as a hypothesis and check it against a later or otherwise held-out sample. Looking at many combinations raises the chance of finding an impressive accident.

Report absolute counts and outcome sizes alongside averages. Explain whether a few trades dominate the result and whether the same definitions were used across groups. There is no universal trade-count threshold that turns a descriptive statistic into certainty. Better documentation makes a limited finding useful without promoting it into a guarantee about the next session.

Sources and further reading

Related guides