How Many Trades to Evaluate a Strategy?

How many trades to evaluate a strategy? Treat 30 trades as the minimum for a first look, 100 as the point where a trend starts to show, and 200 or more before you trust an average or change anything big. Those are rules of thumb, not laws. The real answer depends on how noisy your results are, and noisy results need more trades.

This guide is for futures and prop firm traders at any stage, from a first evaluation to a funded account. Every number below is hypothetical, in USD or R, and the math is our own, built to show the method. Nothing here promises profit. The goal is a way to judge your own sample honestly.

How many trades do you need to evaluate a strategy?

You need enough trades that luck stops dominating the result. For most futures strategies that means roughly 100 to 250 trades for a fair read, with 30 as a bare minimum and more for low win rate or mixed setups. Search results disagree on a single number, and most are trading blogs and forums, so use the ranges as guides.

Here is how the common benchmarks line up.

Your own results will not match a neat table. The sections below show why, with numbers you can reproduce.

Why is 30 trades not enough to prove an edge?

Thirty trades is enough to see whether you followed your plan, but not enough to prove the plan makes money. Results from a small sample swing widely because every trade is a coin flip with uneven payouts. A skilled trader and an unskilled one can look identical over 30 trades.

Start with win rate. Say your true win rate is 55%. The usual 95% range for the win rate you will actually observe is about plus or minus 1.96 times the square root of (0.55 x 0.45 / n), where n is the number of trades.

This is a simple approximation that treats trades as independent, and real markets are messier. Even so, the message holds. Over 20 trades, a trader whose real win rate is 55% can easily print 40% or 70%. If you judged yourself on those 20 trades, you would be judging noise.

How big does the sample need to be for average R?

Win rate is only half the story, so test the number that actually pays you: average R per trade, also called expectancy. If R is new to you, see how to calculate R-multiple in trading first. Use a hypothetical strategy and put real numbers on it.

Suppose a strategy wins 40% of the time. Winners make +2R and losers lose -1R. Expectancy is 0.4 x 2 - 0.6 x 1 = +0.20R per trade. That is a real edge by construction. The standard deviation of a single trade is about 1.47R, so the uncertainty in your average shrinks with the square root of the trade count.

Read the last column slowly. This strategy truly makes +0.20R per trade, and over 20 trades it still shows a losing average about 27% of the time. Over 50 trades, about one time in six. Only near 200 trades does the uncertainty shrink to roughly the size of the edge itself.

That is the reason a quoted figure like "100 trades" feels thin to anyone who runs the numbers. A strong edge, such as +0.5R per trade, needs fewer trades to confirm. A thin edge needs many more. You do not know your true edge, which is why you collect data.

Why do streaks fool small samples?

Streaks fool small samples because a normal losing run looks like a broken strategy. If you lose 60% of your trades, an eight-trade losing streak is not rare when you take enough trades. It is part of the shape of the strategy.

Using a standard run calculation for a 60% loss rate (our own math, not market data):

If you ditch a sound strategy the first time it loses eight in a row, you will often be quitting at the exact moment the sample is still too small to tell you anything. The longer losing-streak math is in what to do after a losing streak in futures.

How long does it take to reach 100 or 200 trades?

It depends on how often you trade, and for many traders the answer is months. That matters, because a sample spread over months can cross different market conditions, and a changed strategy resets the count.

Using a rough 21 trading days a month, here is a hypothetical calendar.

You cannot shortcut this by trading more. Extra trades from overtrading add noise and fees and often dilute the edge. If a slow strategy needs a year of data, trade it at small size and be patient.

Do you need more trades if you test several setups?

Yes. Every setup, symbol or time window you split out needs its own sample, and splitting divides your count fast. A 200-trade log is one decent sample for the whole strategy, but only 50 trades each if you run four setups, and each of those is back in the noisy zone.

Take a hypothetical 120-trade log across four setups, 30 trades each. At the same 1.47R spread per trade, the 95% uncertainty on each setup's average is about plus or minus 0.54R. If setup C shows +0.30R and setup D shows -0.20R, you cannot say C is better than D. The difference is smaller than the noise.

This is the trap in every "my best hour" and "my worst day" analysis. Slice a small log enough ways and something always looks great and something always looks terrible. Treat the results as leads, then collect more trades or run a forward test before acting.

Judge the whole strategy on the pooled sample first, and split only once you have a few hundred trades. The same goes for reviewing your trades each week: a week is a lead, not a verdict.

What counts as a trade in your sample?

Only trades you took the way you planned should count toward evaluating the strategy. A trade where you moved the stop, chased the entry or doubled size is a different strategy: yours on a bad day. Mixing the two hides what you are trying to measure.

This is where behavior changes the sample. Suppose a hypothetical 100-trade month splits as follows.

All 100 trades together look like a thin +0.09R edge. The 80 plan-following trades look healthier, though 80 is still a small sample. The plan-breaking group is the real story: 20 trades that cost 11R, more than half of what the clean trades earned. That is the cost of a habit, and it comes from you, not the market.

How do you read a small sample without fooling yourself?

Use small samples to check process, not to size up edge. With fewer than 100 trades, ask whether you followed your rules, whether your risk per trade stayed constant and whether your journal is complete. Those questions have clear answers on 20 trades. Edge questions do not.

A short checklist for small-sample periods:

1. Fix the rules first. Write the entry, stop, target and size rule before you count anything. Changing rules mid-sample restarts the count.
2. Keep risk constant. Trade small and the same size so each trade counts the same.
3. Log every trade. A sample missing the trades you are embarrassed by is a biased sample.
4. Tag plan-following and plan-breaking trades. Evaluate them separately.
5. Check the best-trade test. Recalculate your average without your single biggest winner. If the edge disappears, it was one trade. The profit factor guide shows this method.
6. Set a review date. Decide in advance, for example at 100 trades, when you will look and what would change your mind.

What can Traderilla show you here?

Traderilla gives you the raw material: your trade count, average dollars per trade (expectancy) and profit factor on the Analytics page, plus the five automatic behavior flags on every trade you log or import. Those are revenge trades, cool-down violations, overtrading, consecutive losses and oversized positions.

It does not calculate confidence ranges, tell you whether a sample is big enough or tag your trades as plan-following, so the tags and the statistics above are yours to keep. The app warns you but cannot block real orders. Import is by CSV upload, one row per completed trade.

One plan limit matters for sample size. Free logs 30 trades a month and shows the last 30 days of history on screen, so a long sample needs a paid plan, which has unlimited trades and full history. Older Free trades are kept and come back on upgrade. Setup, day and hour breakdowns are also on Pro and Elite only, so on Free you see overall numbers, not slices, which suits small samples anyway.

Key takeaways

• Thirty trades checks your process. It cannot prove an edge.
• In the hypothetical example, a real +0.20R edge still showed a losing average about 27% of the time over 20 trades, and about 3% of the time over 200.
• A 55% true win rate can print anywhere from about 33% to 77% over 20 trades.
• Normal losing streaks fool small samples. Eight losses in a row happens in about half of 100-trade samples when you lose 60% of the time.
• Splitting a log by setup, day or hour divides your sample size. Treat slices as leads.
• Count only trades you took by the plan when judging the strategy, and price the plan-breaking trades separately.
• Decide your review point in advance, and change one thing at a time.

FAQ

Is 100 trades enough to test a trading strategy?

It is enough for a rough read and to catch big flaws, but not to prove a small edge. In the hypothetical example above, 100 trades still left an uncertainty of about plus or minus 0.29R around a true +0.20R edge. Treat 100 as a checkpoint and keep collecting data.

Is 30 trades a good sample size for trading?

It is a common minimum, not a good sample. Thirty trades shows whether you followed your rules, but results swing widely at that size. Use 30 to check execution, then keep going before you trust win rate, average R or profit factor.

How many trades per setup do I need?

Each setup needs its own sample, so aim for the same 100 to 200 trades per setup if you want to compare them. If you run four setups, that means several hundred trades overall. With fewer, treat differences between setups as leads, not conclusions.

How many trades do I need to judge a prop firm evaluation strategy?

An evaluation often lasts only a few weeks, so the sample you get inside it is small. Use it to check process and risk control, not to prove edge. Rules and targets differ by firm and change, so read your firm's current rules on its own site.

Does the number of trades matter more than the number of months?

Trade count matters for the statistics, but the months matter for market conditions. A 150-trade sample from one calm month tells you less than 150 trades across calm and volatile periods. Aim for both: enough trades, spread over different conditions.

Should I change my strategy after a bad week?

Rarely. One week is a small sample and a normal losing streak can fit inside it. Check whether you followed your rules first. If the plan-following trades are fine, collect more data, and only change one thing at a time.

Traderilla imports your Tradovate or NinjaTrader CSV exports, flags mistakes automatically and shows what your most expensive habit costs you in dollars, even on the free plan (30 trades a month, no card). Paid plans show the cost of every habit, or try Elite free for 7 days (then $59/month unless you cancel).