How many trades do you need for a reliable forex indicator backtest?

In this guide

Why the count matters
Practical guidelines
Count is not the only issue
What to measure
Related

Aim for at least 100 trades as a bare minimum, and 200 to 300 or more if you want to lean on the numbers. Below about 30 trades, results are mostly noise. Even at 100, the plausible range around your win rate or profit factor is wide, so treat the result as a screening tool and not as proof.

Why the count matters

Think of a coin flip. Ten flips can easily give you seven heads. A small sample of trades behaves the same way: a lucky streak or a single trending month can make a mediocre indicator look excellent. As a rough rule of thumb, the uncertainty around a measured win rate shrinks only with the square root of the sample size, so doubling your trades does not halve the doubt.

Practical guidelines

Under 30 trades: anecdote. Do not draw conclusions.
30 to 100 trades: useful for ruling out obvious failures, not for confirming a winner.
100 to 300 trades: a reasonable working sample for a single pair and timeframe.
300+ trades: better, especially if they span different market conditions.

These are rules of thumb, not statistical guarantees. Strategies with a low win rate and large winners need more trades than high-win-rate ones, because a few big trades can dominate the totals.

Count is not the only issue

Spread the sample across conditions

Three hundred trades taken entirely in one strong trend tell you little. Make sure your data includes ranging, trending and high-volatility periods, and ideally more than one pair.

Check for repainting first

A large sample is worthless if the indicator redraws past signals. Rule that out before you count anything. See Repainting Indicators Explained.

Include costs

Spread, commission and slippage can turn a thin edge into a loss. Record them on every trade.

Keep out-of-sample data

If you tweak settings until the backtest looks good, you have fitted the past. Reserve a separate period you never optimized on, then confirm with demo and forward testing.

What to measure

Log win rate, average win and loss, profit factor and maximum drawdown. For context on what the numbers should look like, read what is a realistic win rate and profit factor. A backtest cannot promise future results, and nothing here is financial advice. Only risk money you can afford to lose.