A 90% win rate system can lose money. A 40% win rate system can make a fortune. If that sounds wrong, this article is for you, because these three numbers only mean something together.
Win rate: the most overrated number
Win rate is the percentage of trades that close profitable. It is also the easiest number to manipulate: take profits at 5 pips and use 100-pip stops and you will win constantly, right up until the stops erase everything. Our two systems demonstrate the range: the Breakout System wins 57.2% of trades, the Mean Reversion System wins 72.4%. Neither number alone says which makes more money.
Profit factor: the quality number
Profit factor is gross profits divided by gross losses. A profit factor of 1.0 is breakeven. Above 1.5 is solid, above 2.0 is strong. This number captures what win rate hides: the size relationship between winners and losers. Our live breakout account currently runs a profit factor of 1.60, meaning every dollar lost was answered by $1.60 won.
Expectancy: the money number
Expectancy is the average profit per trade, combining both previous numbers: (win rate x average win) minus (loss rate x average loss). This is the number that actually compounds. A system with positive expectancy makes money over enough trades regardless of its win rate profile. Two examples with identical expectancy:
| Breakout style | Mean reversion style | |
|---|---|---|
| Win rate | 57% | 72% |
| Avg win | 0.54% | 0.26% |
| Avg loss | -0.50% | -0.43% |
| Character | Fewer, larger wins | Many small wins, rarer larger losses |
How to use the trio when evaluating an EA
- Ignore any marketing that leads with win rate alone. Especially above 85%, which usually signals martingale mechanics or tiny-target scalping.
- Ask for profit factor and trade count together. A 2.5 profit factor over 80 trades is noise. Over 4,000 trades it is a system.
- Compute expectancy per trade and multiply by annual trade frequency. That is the honest return engine, before compounding.
All three numbers for both of our systems are in the full backtest reports, trade by trade, included with every purchase.
The sample size caveat on every statistic
All three numbers are estimates whose reliability scales with trade count, and the error bars are wider than intuition suggests. A genuine 57% win rate system will, by pure chance, show anywhere from roughly 47% to 67% over any given 100-trade window. Profit factor is even noisier because a couple of outlier trades swing it hard: the same underlying system can print a 1.4 and a 2.4 in adjacent quarters. This has two practical consequences. When evaluating a product: statistics quoted from a few hundred trades deserve heavy skepticism, and our preference for publishing the full 4,542-trade and 1,438-trade datasets is precisely about shrinking those error bars to something meaningful. When monitoring your own live account: a 60-trade cold streak that drops the rolling win rate to 51% is almost always variance, not decay, and reacting to it is how traders destroy working systems. The discipline is to predefine the sample size (we use 100+ trades) below which no statistic is allowed to trigger a decision.
A worked expectancy calculation, start to finish
To make the money number concrete, here is the full arithmetic on the breakout system's backtest statistics. Win rate 57.2%, so loss rate 42.8%. Average winner 0.54% of account, average loser 0.50%. Expectancy per trade: (0.572 x 0.54) minus (0.428 x 0.50) = 0.309 minus 0.214 = roughly +0.095% of account per trade. Around 0.1% per trade sounds almost insultingly small, which is exactly the point most buyers miss: the system takes several hundred trades per year, and 0.095% compounded across that frequency, with position sizes growing as equity grows, is what produces the double-digit annual figures. Edges in liquid markets are thin by nature; sustainable systems win by expressing a thin edge relentlessly at controlled risk, not by finding a thick one. Any product whose implied per-trade expectancy looks fat should trigger the curve-fitting questions from elsewhere on this blog, because thick edges in majors get arbitraged out of existence, while thin, structural, boring ones survive.