How many trades do you need to test a strategy?
There is no magic number of trades that proves a strategy works. Twenty is a first look. A hundred is a more useful check. Neither replaces clear rules and honest records.
The short answer
Five wins do not prove an edge. Neither do 20 trades. Aim to log at least 30 trades as an early check and 100 or more as a stronger review, but do not treat 100 as a pass mark. The real test asks whether the same written rules held up across different conditions, after costs, without looking ahead.
You need two kinds of evidence: enough trades to see the range of outcomes, and a test you did not quietly change halfway through. A long, biased test is worse than a short, honest one.
What 20 trades can tell you
Say you make 20 hypothetical trades with 10 wins. Your observed win rate is 10 / 20, or 50%. It feels like a solid number. But the number can move a lot when just a few more trades arrive.
A simple way to show that uncertainty is a 95% Wilson interval. It is a statistical range for a win rate estimated from wins and losses. The range assumes the trades are independent and come from one stable process. Markets rarely meet those assumptions perfectly, so use it as a warning against false certainty, not as a guarantee.
Here is the range for a made-up test that wins exactly half the time. Each row uses the same 50% observed win rate.
| Trades | Wins | Observed rate | Approximate 95% range |
|---|---|---|---|
| 20 | 10 | 50% | 30%–70% |
| 30 | 15 | 50% | 33%–67% |
| 50 | 25 | 50% | 37%–63% |
| 100 | 50 | 50% | 40%–60% |
| 200 | 100 | 50% | 43%–57% |
The table is not a promised result. It shows how a small sample leaves more room for luck. The full calculator uses the Wilson method and displays one decimal place.
Use the free win-rate range calculator with your own trade count and wins. Then check the expectancy calculator with average win and loss sizes. A method can win most trades and still lose money when the losses are much bigger.
Count the trades you did not want to count
Imagine two people testing the same idea. Both write down 50 winners and 50 losers.
One recorded every entry and exit as the chart played out. The other skipped three bad entries after seeing the next candle and changed the stop twice. Their spreadsheet totals are identical. Their evidence is not.
Before the first trade, write down:
- What qualifies. Market, chart time frame and the exact setup.
- What cancels it. The conditions that make you skip the trade.
- Where you enter. A clear trigger, not "near support."
- Where you exit. Stop, target and any time-based rule.
- Which costs you count. Spread, commission and slippage assumptions.
- When you will review. Choose a batch size before seeing the scores.
If you change one rule, mark the date and start a new version of the test. Do not blend the two groups and call them one strategy.
A two-stage test plan
The following is a practical study plan, not a statistical pass/fail rule.
| Stage | Work | What you decide |
|---|---|---|
| First 30 trades | Run one fixed setup; record every attempt | Are the rules clear enough that you can follow them? |
| Next 70 trades | Keep the same rules; include other market conditions | Are the process and results still consistent? |
| Beyond 100 | Repeat on later dates and new periods | Does the pattern hold outside the original sample? |
Thirty trades are not enough to estimate a narrow win rate. Their job is to expose broken instructions: "I could not tell if that was an entry." Fix that, call the new rules version 2, and test again.
The spreadsheet that makes a test believable
Use one row for each trade. Include the ones that lost and the ones you wish you had skipped. This is a small hypothetical log:
| Trade | Setup version | Market state | Result (R) | Followed rules? |
|---|---|---|---|---|
| 1 | A | Trending | +2R | Yes |
| 2 | A | Sideways | -1R | Yes |
| 3 | A | Trending | -1R | No |
| 4 | A | Sideways | +1R | Yes |
R means one unit of planned risk. A 1R loss is a loss the size of the amount you chose to risk. In this four-trade example, total R is +1R, but one entry broke the plan. The sample is far too small for a verdict. You can use the free journal CSV to keep the full set.
Count rules followed separately from win rate. If the rules were not followed, that result tests your behavior as much as the strategy.
Three ways a large sample can still fool you
One market mood. A hundred trades in one strong trend are a hundred examples of that trend. Include ranging periods, quiet periods and days when the setup did not appear.
Hindsight. If you can see future candles, every entry looks obvious. In replay, make the call and write it down before showing the next candle. How to backtest a strategy explains that discipline.
Changing the question. You try five stop distances, keep the best, and then report its result as if it were the only plan. That is a search for a good-looking past, not a fresh test. Use a later untouched sample to check the chosen settings.
The decision after the sample
Review the trades in three passes. First, check whether the rules were followed. Second, calculate win rate, average win, average loss and expectancy. Third, compare market conditions. If the results fall apart only in sideways periods, that is something to investigate, not a reason to rewrite every past row.
StructureX members can record practice decisions and review their journal habits. You can also do this for free in a spreadsheet. The numbers matter more than the app that holds them.
For the next step, read how long to paper trade and how to measure trading progress. They answer two different questions: when to move from practice to live, and whether your decision process is getting better.
Questions
Is 30 trades enough to test a strategy?
Thirty trades can show whether your rules are clear and whether your first results are worth investigating. They cannot settle a narrow win-rate estimate. Keep testing across different conditions.
Why do people say 100 trades?
One hundred is a manageable review point with much less sampling noise than 10 or 20. It is a checkpoint, not proof of future results.
What if the first ten trades all win?
Keep the same plan and record the next trades. Ten wins can happen by chance. Do not change your risk or skip the difficult setups because the first ten looked good.
What is more important than win rate?
Average win, average loss, costs and whether you followed the rules. The expectancy calculator combines win rate with the size of wins and losses.
Do paper trades count?
They count as evidence about following rules and testing setups. Simulated fills and real emotions differ from live trading, so keep a separate label for paper and live results.
Follow along