# How Many Trades Before Your Stats Are Reliable

> Trading statistics from small samples are unreliable. Here is what sample sizes are needed for different levels of statistical confidence.

**Tags:** sample-size, statistics, reliability, data-quality
**URL:** https://traderjournal.app/trading-metrics/how-many-trades-before-stats-are-reliable

---


# How Many Trades Before Your Stats Are Reliable

One of the most common mistakes in trading analysis is drawing conclusions from too few trades. A 70% win rate over 10 trades tells you almost nothing. Here is what sample sizes you actually need for reliable statistics.

---

## Why Small Samples Are Misleading

Probability produces streaks. At a true 50% win rate (like a coin flip), consecutive runs of 5 losses in a row happen approximately 3% of the time. That sounds rare, but across a trading year with 500 trades, you expect this to happen about 15 times.

This means: in any 10-trade window, your observed win rate can deviate significantly from your true win rate due to random clustering. A 70% observed win rate over 10 trades could come from a strategy with a true win rate of 45% just by lucky clustering.

---

## Rough Sample Size Guidelines

These are practical guidelines, not precise statistical thresholds:

**For identifying whether a strategy has positive expectancy:**
Minimum 100 trades. At 100 trades, the observed expectancy is within approximately ±1R of the true expectancy with reasonable confidence.

**For reliable win rate estimates:**
100-200 trades. Expect the observed win rate to be within approximately ±5 percentage points of the true rate.

**For profit factor estimates:**
200+ trades. Profit factor is sensitive to outliers, so it stabilizes more slowly than win rate.

**For by-symbol or by-session analysis:**
20-30 trades per category minimum. If you trade EURUSD 100 times and GBPJPY 8 times, your GBPJPY stats are not reliable regardless of what they show.

**For strategy comparison:**
300+ trades per strategy with identical market conditions (same period, same conditions). Comparing two strategies over different periods introduces market condition bias.

---

## Confidence Intervals in Practice

A simple way to think about it: with N trades, your true win rate is probably within approximately 1/sqrt(N) of your observed win rate.

At N=25 trades: uncertainty of ±1/5 = ±20 percentage points
At N=100 trades: uncertainty of ±1/10 = ±10 percentage points
At N=400 trades: uncertainty of ±1/20 = ±5 percentage points
At N=1600 trades: uncertainty of ±1/40 = ±2.5 percentage points

These are rough approximations. The key insight: most retail traders with 50-100 trades are making decisions based on statistics with ±10-15% uncertainty. That is enough uncertainty to make a losing strategy look like a winning one.

---

## What to Do While Data Accumulates

During the early phase (under 100 trades), focus on:
- Process quality (following rules, consistent execution)
- Data collection (consistent tagging, complete notes)
- Forming hypotheses about your edge rather than drawing conclusions

After 100 trades, begin tentative analysis but hold conclusions loosely.

After 200 trades, your statistics are starting to mean something. Patterns that appear across 200 trades are more likely to be real than ones that appeared at 50.

After 500+ trades, you have enough data to make confident strategy decisions based on your journal analytics.

---

This is why the 90-day commitment to journaling is so important. It is not enough time to prove your edge but it is enough to begin seeing patterns. The long-term player thinks in years of data, not weeks.

Download Trader Journal at android.traderjournal.app or ios.traderjournal.app and start accumulating reliable data.