How Many Trades Do You Need Before Your Statistics Mean Anything?

Ten winning trades in a row feels like proof. It is not. With small samples, luck can easily look like skill and skill can look like failure. Knowing how many trades you need before trusting a statistic saves you from abandoning good strategies and sticking with bad ones.

How much a win rate can vary

The uncertainty in a win rate shrinks with the square root of the number of trades. A rough 95% range for an observed win rate p over n trades is:

p ± 2 × √(p × (1 − p) ÷ n)

For an observed win rate of 55%:

After 20 trades, a "55% strategy" could plausibly be a 35% strategy or a 75% one. Even after 100 trades, the range is ten points wide either way. This follows from the law of large numbers: averages settle down only as the sample grows.

Practical minimums

There is no magic number, but as a rough guide:

• 30 trades: enough for a first look, not for conclusions.
• 100 trades: a reasonable basis for evaluating one setup.
• 200 or more: much more confidence, especially for low-win-rate strategies whose results depend on occasional large winners.

Each breakdown, such as by setup, hour or weekday, divides your sample. A setup with 15 trades has 15 trades, even if your journal has 500.

Monte Carlo simulation

A practical way to see the range of outcomes is Monte Carlo simulation:

1. Take your list of trade results in R.
2. Randomly reshuffle them, or sample with replacement, thousands of times.
3. Look at the range of final results and drawdowns.

This shows how different your results could have been with the same edge, and how deep a drawdown you should be prepared for. See risk of ruin.

Other traps

• Changing the strategy resets the sample.
• Changing markets. A sample from one regime may not apply to another.
• Cherry-picking periods. Look at all your data, not your best month.

What to do in the meantime

While your sample is small, focus on process: follow your plan, record everything and keep risk small. The statistics will become meaningful with time.

In TradeThink

Every metric is shown with the number of trades it is based on, so you can tell a solid result from a thin one. Filter by setup in the reports and check the trade count before drawing conclusions. The free Kelly's Simulator shows how much random paths vary for a given edge.