How Many Trades Does It Take To Evaluate A Strategy

Introduction

A trader tests a new strategy. After 10 trades, he has 6 winners and 4 losers. 60% win rate. Promising. He increases size. At trade 15, two losses back-to-back. His rate drops to 53%. He panics. He abandons the strategy and tries another.

Six months later, he's on his seventh strategy. None got a real chance to work. He thinks the problem is the strategies. The truth? It's the sample size.

With 10 trades, you see nothing. With 30 trades, you start seeing patterns. With 100 trades, you have reliable evaluation. But many traders don't wait. They change after 15, after 20, after 30. And they abandon genuinely profitable strategies from statistical impatience.

Why Small Samples Lead To False Conclusions

The fundamental problem: variance.

Imagine a strategy with true expectancy of +0.5% per trade. Over 1000 trades, that's +5%. Excellent. But over 10 trades? You could easily make -3% from pure chance.

Here's how variance works:

True strategy: 55% win rate, average win 1.2%, average loss 1%.

True expectancy = (55% × 1.2%) - (45% × 1%) = 0.66% - 0.45% = +0.21% per trade.

Profitable strategy. But watch what can happen over 20 trades:

Scenario 1: 13 winners, 7 losers. Total profit: +7.1%. Win rate: 65%.

Scenario 2: 10 winners, 10 losers. Total profit: +0.2%. Win rate: 50%.

Scenario 3: 9 winners, 11 losers. Total profit: -2.2%. Win rate: 45%.

Same strategy. All three scenarios are statistically plausible. A trader testing 20 trades and hitting scenario 3 will think his strategy doesn't work. He'll abandon it. He'll be wrong.

The Critical Statistical Threshold: Where Variance Begins To Quiet Down

There's a point where the sample becomes large enough that variance calms and reality emerges. Statistically, that's around 30 to 50 trades for strategies with typical win rates (40-60%).

Here's why:

With 10 trades, variance can mask or exaggerate your true results by 2x or 0.5x.

With 30 trades, that variance reduces significantly. You start seeing real expectancy.

With 100 trades, you have excellent picture of true strategy behavior.

With 500 trades, you're nearly certain.

Numerical example:

A strategy with true expectancy +0.5% per trade.

• After 10 trades: result can vary from -1.2% to +2.1% (huge range).
• After 50 trades: result can vary from -0.8% to +1.8%.
• After 200 trades: result can vary from -0.2% to +1.2%.
• After 500 trades: result can vary from +0.1% to +0.9%.

See? At 500 trades, variance is nearly eliminated. You're certain of real strategy behavior.

The Three Evaluation Levels

Level 1: Exploration (5 to 15 trades)

You're testing an idea. You have zero certainty. Variance is enormous. Conclusions here are highly unreliable.

Use: identify major bugs (crashes, coding errors, missing configs).

Conclusion: "Strategy operates without obvious errors." Full stop.

Level 2: Provisional evaluation (30 to 50 trades)

You're starting to see real patterns. Not absolute certainty, but a direction. Variance still affects results, but it's manageable.

Use: identify if strategy has positive or negative expectancy. Make major adjustments.

Conclusion: "Strategy appears positive (or negative) on preliminary trend."

Level 3: Solid evaluation (100+ trades)

You now have good picture of strategy. Variance exists, but you can distinguish signal from noise.

Use: fine optimization, identify cycles, correlate with market context.

Conclusion: "Strategy has stable expectancy of X% and win rate of Y%."

Expectancy vs Win Rate: Why Sample Size Matters More For One

If you measure win rate, small samples create huge swings.

50% of 20 trades = 10 winners, 10 losers.
50% of 21 trades = 10.5 winners (impossible), so 10 or 11.

Result jumps between 47.6% and 52.4% with ONE single trade.

But expectancy? It's more stable because it combines win rate AND magnitude. You might have slightly wounded expectancy from one large loser, but your overall estimate stays more consistent than with win rate alone.

For this reason, expectancy is better metric for evaluating strategies. But even expectancy needs sufficient sample.

Common Mistakes: How Traders Abandon Profitable Strategies

Mistake 1: Evaluating on too few trades.

A trader tests 15 trades, gets -2%, abandons. Classic error. With 15 trades, you can easily get negative variance even if your strategy is truly positive.

Mistake 2: Changing the strategy instead of waiting for it.

After 20 trades, strategy is slightly negative. Instead of waiting 80 more trades, the trader makes a "small tweak." Then another. Eventually, it's a different strategy. He never evaluated the original.

Mistake 3: Confusing variance with systemic weakness.

20 trades, 45% win rate. Trader thinks "my strategy is broken." But 45% on 20 trades can be perfectly normal for a true 52% strategy. It's just variance.

Mistake 4: Not counting backtesting trades.

A trader backtests 100 trades, then goes live. After 50 live trades, he evaluates. But he should count 100 + 50 = 150 total trades. Not doing so underestimates sample size.

Mistake 5: Stopping at the worst moment.

A trader tests 40 trades. At 30, he's at maximum drawdown. At 40, he's recovered. If he'd stopped at 30, he'd have thought the strategy doesn't work. Evaluation horizon is unfortunately "when you stop," not "when you planned."

Risk Discipline: Managing Variance During Evaluation

Waiting 100 trades to evaluate is good. But you must manage risk during the wait.

A true strategy but unlucky can put you in 15-20% drawdown during first 50 trades. If you'd risked 5% per trade, you might have lost 50% of capital before discovering the strategy works.

Solution: risk less during evaluation. Risk 1% per trade, not 2-3%. Let variance play its role without ruining you. If the strategy survives 100 trades at 1% risk, you know it can handle variance. Then increase to 2%.

Best Practices: How To Evaluate Correctly

1. Set a minimum number BEFORE starting.

Not "I'll do 50 trades and see." Not flexible. You fix a number: 80 trades. You don't change evaluation before 80.

Why 80? Because 50-100 is where variance becomes manageable.

2. Document your evaluation metric BEFORE starting.

Will you evaluate on: expectancy, win rate, or P&L?

If you evaluate on win rate, that's an error (see expectancy article). Use expectancy. It's the true metric.

3. Track progress every 10 trades.

You don't decide every trade. But every 10 trades, you note: expectancy, win rate, P&L. This shows you trend without creating false emotional signals.

4. Separate market context from strategy performance.

You test 50 trades in strong trending market. You get 65% win rate. Nice stat. But you don't know how your strategy behaves in ranging markets. Test it in different contexts before final evaluation.

5. Mentally prepare for variance.

First 30 trades will be chaotic. Accept it. It's normal. Don't change your strategy. It's variance. Keep going.

6. After 100 trades, you can conclude.

If expectancy is positive: strategy probably works. Increase risk progressively.

If expectancy is negative: strategy probably doesn't. Stop or revise.

Conclusion

The number of trades needed to evaluate a strategy isn't arbitrary. It's a statistical question. 30-50 trades for first direction. 100+ trades for reasonable certainty.

Many traders abandon profitable strategies because they evaluate on 15-20 trades. They confuse variance with weakness. Conversely, some traders stay stuck with losing strategies because they wait 500 trades "to be sure."

The balance: enough trades for variance to calm (100), but not so many you destroy your capital first (not 1000).

Set your minimum number. Document your metric. Accept variance. Give your strategy a real chance before judging it.

A strategy isn't success or failure after 20 trades. It's promising or doubtful after 100.