Win Rate Is Not a Performance Metric — Expectancy Is
The most common mistake new EA traders make is judging an EA by its win rate. It feels intuitive — more wins means more profit, right? But this logic collapses entirely when you account for the size of the wins versus the size of the losses.
Consider two EAs running on XAUUSD. EA-A wins 70% of its trades but the average win is 4 pips and the average loss is 12 pips. EA-B wins 45% of its trades but the average win is 15 pips and the average loss is 8 pips. Which one would you want?
EA-A: (0.70 × 4) - (0.30 × 12) = 2.8 - 3.6 = -0.8 pips per trade expectancy. Negative — it loses money over time. EA-B: (0.45 × 15) - (0.55 × 8) = 6.75 - 4.4 = +2.35 pips per trade expectancy. Positive — it grows the account over time.
EA-A sounds better because it wins more often. EA-B is actually far superior because it has positive expectancy. This is the fundamental disconnect that allows marketing for low-quality EAs to target people with impressive-sounding win rates. Before evaluating any EA for purchase, demand expectancy data, not win rate headlines.
The Expectancy Formula Explained
Expectancy is expressed in R units, where 1R equals the amount you risk per trade. If you risk $50 per trade, then 1R = $50.
Expectancy Formula
E = (WR × Avg Win) - (LR × Avg Loss)
WR = Win Rate (decimal) | LR = 1 - WR
Avg Win and Avg Loss both in R units
If 1:1 R:R: E = (WR × 1) - ((1-WR) × 1) = 2WR - 1
Break-even at 1:1 R:R requires exactly 50% win rate
Trend follower
Win rate: 40%
R:R: 1:2
Expectancy: +0.20R
Profit factor: 1.33
Gold scalper (typical)
Win rate: 55%
R:R: 1:1.5
Expectancy: +0.38R
Profit factor: 1.83
High WR, poor R:R
Win rate: 70%
R:R: 1:0.5
Expectancy: +0.05R
Profit factor: 1.17
The third example — 70% win rate with 1:0.5 R:R — shows exactly why high win rate alone is misleading. At 0.70 × 0.5 = 0.35 minus 0.30 × 1 = 0.30, the expectancy is just +0.05R per trade. Marginal profitability at best, and any increase in spread or slippage costs can turn it negative.
What Real Win Rates Look Like for Gold Scalping EAs
After reviewing hundreds of backtest reports and live account statements across XAUUSD scalping strategies, the realistic range for legitimate gold scalping EAs is 45%–65% win rate. Here is what drives where an EA lands in that range.
| Strategy Type | Win Rate Range | Typical R:R | Expectancy Range |
|---|---|---|---|
| Breakout scalper (M1) | 48–58% | 1:1.2–1:1.8 | +0.08–+0.22R |
| Session open sniper | 52–65% | 1:1.0–1:1.5 | +0.07–+0.25R |
| Low-frequency sniper | 55–70% | 1:0.8–1:1.2 | +0.04–+0.22R |
| Semi-manual scalper | 50–68% | 1:1.0–1:2.0 | +0.00–+0.36R |
| Martingale disguised as EA | 88–97% | 1:0.05–1:0.2 | Negative long-term |
The martingale row illustrates the pattern clearly. Very high win rate with microscopic average wins and catastrophic occasional losses. The expectancy is not just negative — it is a time bomb. This is explored in more depth in the article on comparing EA safety profiles.
Win Rate, Drawdown, and the Psychology of Losing Streaks
One of the most important (and least discussed) consequences of win rate is its impact on losing streak probability. Every win rate implies a distribution of consecutive losses that will occur in any sufficiently long trading period.
Expected maximum losing streak by win rate (over 200 trades)
Up to 12 consecutive losses
Extreme but statistically expected
Up to 9 consecutive losses
Occurs roughly twice per year at 80 trades/month
Up to 8 consecutive losses
Normal — expect this to happen occasionally
Up to 6 consecutive losses
Shorter streaks but still inevitable
This table explains why many EA users shut down their EA at exactly the wrong time — after a streak of 6–8 losses, convinced the system has stopped working. Statistically, that streak was expected and does not indicate any change in the underlying edge. The EA that just had 8 losses has the same positive expectancy it had before the streak — and the next 8 trades are independent of the previous ones.
The practical implication: size your positions to survive the maximum expected losing streak. If you expect up to 10 consecutive losses and each loss is 2% of account, you need to survive 20% drawdown without your account going below your broker's margin limit. This is one of the core common mistakes including chasing win rate that experienced traders warn about.
The 95% Win Rate Red Flag — How to Identify Martingale in Disguise
The most dangerous EAs in the market are not obviously bad — they look spectacular in their marketing materials. A 95% win rate across 1,000 trades sounds extraordinary. The question is: what do the 5% losing trades look like?
If an EA shows 950 winning trades averaging +4 pips and 50 losing trades averaging -78 pips, the math tells the story. Average gain per trade: (950/1000 × 4) = +3.8 pips. Average loss per trade: (50/1000 × 78) = -3.9 pips. Net: -0.1 pips per trade. Negative expectancy despite a 95% win rate — and the losses are 19.5x the size of the wins.
How to spot this: download the full trade history from any backtest report and look at the 10 largest single losing trades. If any single losing trade exceeds 10x the average winning trade size, you are looking at martingale or grid mechanics, not a genuine edge. Legitimate scalping EAs have relatively consistent loss sizes because they use fixed stop losses.
The portfolio win rate diversification approach — running multiple EAs with genuinely different strategies — is a more robust way to achieve high overall account win rates without the hidden risks of a single high-win-rate system.
Profit Factor: The Single Most Useful Performance Number
If you had to pick one number from a backtest report, it should be profit factor. Profit factor = total gross profit divided by total gross loss. A profit factor of 1.5 means for every $1.00 lost in losing trades, the system made $1.50 in winning trades — a 50% net return over the period tested.
PF: Below 1.0
Losing systemCosts more in losses than it earns in wins. Do not run live under any circumstances.
PF: 1.0–1.2
MarginalMay be profitable but has thin edge easily eroded by spread changes or market condition shifts.
PF: 1.3–1.8
Good — deploy carefullyRealistic range for legitimate gold scalping EAs. Enough margin above 1.0 to survive real-world conditions.
PF: 1.8–2.5
Excellent — verify sample sizeStrong edge. Ensure the backtest covers 500+ trades across multiple market regimes before trusting this figure.
PF: Above 2.5
Suspicious — check for overfittingExtremely high profit factors on historical data often indicate curve-fitting. Test forward on unseen data.
PF: Above 5.0
Almost certainly overfittedReal edge does not produce these numbers at scale across diverse market conditions. Be sceptical.
How to Evaluate an EA's Win Rate Before Buying
When evaluating any gold EA, here is the complete checklist for win rate and performance analysis:
Request the full backtest report (not a screenshot — a full MT5 report in HTML format) showing individual trades.
Calculate expectancy using the average win size and average loss size from the report, not the stated win rate.
Check profit factor. Reject anything below 1.25 or above 3.0 without explanation.
Verify the backtest period covers at least 2–3 years and multiple market conditions.
Check the largest single losing trade. If it exceeds 10x the average win, investigate for martingale.
Request a live account statement (myfxbook verified) in addition to backtests — backtests can be manipulated.
Run the EA on a demo account for at least 30 trades before live deployment.