Treating a high win rate as a low-risk strategy
The most reliable way to raise a win rate is to hold losers open until they come back, which converts closed losses into floating drawdown. The count improves while the actual risk goes up.
Win rate
The percentage of trades that close at a profit, calculated as winning trades divided by total closed trades. Must be evaluated alongside risk/reward ratio to be meaningful.
also: win percentage, accuracy, hit rate
Updated
In plain English
How often the strategy was right, counted as a share of all its closed trades. It says nothing about how much it made when it was right or lost when it was wrong, which is why it cannot be read on its own.
Why it matters
Win rate is the statistic most often put on an EA's sales page and the one that carries the least information by itself. It matters because it is half of the expectancy calculation — and because knowing that it is only half is what stops a 90% figure from looking like an edge.
- It sets the shape of the strategy, not its quality. A high win rate implies small wins against occasional large losses; a low one implies the reverse. Neither is better until the size of the wins and losses is known.
- It predicts how the strategy feels to run. A 40% win rate means losing runs of five or six are routine, which is a psychological fact about the deployment as much as a statistical one.
- It is the input people most often mis-supply to a position-sizing or risk-of-ruin calculation, where a figure taken from a backtest with unrealistic fills produces a comfortable answer to the wrong question.
- It barely ranks the catalogue it is printed on. Sort the fourteen EAs published here by win rate, then by profit factor, and the two orders come out almost unrelated — a rank correlation of −0.05. The number that sells the product says next to nothing about whether the product made money.
How it is calculated
Win rate % = Winning trades ÷ Total closed trades × 100 - Winning trades
- Trades closed with a profit after costs, in the same window as the denominator.
- Total closed trades
- Every trade that reached a close in the window, wins and losses together.
Only closed trades count. Breakeven trades have to be assigned explicitly — counting them as wins can move the figure by several points on a scalping strategy, so the convention must be stated alongside the number.
What counts as high or low
- typically 35–55% Trend-following EAs — few large wins, many small losses
- typically 60–75% Scalping EAs — many small wins, slightly smaller losses
- typically 70–85% Grid and range EAs — small wins bought with a rare large loss
- over 90% Above 90% on any EA — the losses are being deferred, not avoided
Read it together with
No single number describes a strategy. These metrics change how this one should be read.
Example
Two EAs are advertised on their win rate. Multiplying each rate by what it wins and loses shows which one actually has an edge.
- EA A — win rate
- 80%
- EA A — average win / average loss
- $25 / $150
- Expectancy = 0.80 × 25 − 0.20 × 150 = −$10 per trade.
- EA B — win rate
- 40%
- EA B — average win / average loss
- $300 / $100
- Expectancy = 0.40 × 300 − 0.60 × 100 = +$60 per trade.
Win rate ranked these two the wrong way round because it is a count, and only the sizes make it money.
Calculation 0.80 × 25 − 0.20 × 150 = −10 · 0.40 × 300 − 0.60 × 100 = +60
Result The 40% strategy is the profitable one
How to interpret it
Read win rate as one term of a product, never as a score. The question it answers is only useful once the average win and average loss are on the table.
| Range | What it means |
|---|---|
| Quoted without an average win / loss figure | Uninterpretable. The same win rate is consistent with a strongly profitable and a strongly losing strategy. |
| Above 85% on a strategy that averages down | The count is being kept high by never closing losers. The losses exist as floating drawdown rather than as trades. |
| Consistent between in-sample and out-of-sample | A good sign — win rate is one of the more stable statistics, so a large shift between windows usually means the exit rules were fitted. |
| Measured over several hundred trades | Reliable enough to plan around. Below about a hundred trades the figure moves several points on chance alone. |
| Taken from a short demo or forward run | Descriptive of the run, not of the strategy. A few weeks of trading rarely contains the market condition the exit rules were designed to survive, so the figure tends to sit above the tested one. |
- Pair it with the risk-reward ratio at entry. A 54% win rate on 2:1 reward-to-risk is a better business than an 80% win rate on 1:3.
- Use it to predict losing runs before you deploy. At a 40% win rate, a run of six losses appears regularly over a few hundred trades; if that would make you switch the EA off, the strategy is not deployable at that size.
- Check how breakeven trades were counted before comparing two published figures.
- Expect the payoff to move against the count, and measure by how much. Across the fourteen runs published here the two trade off almost perfectly: rank them by win rate, then by average win against average loss, and the orders invert at a rank correlation of −0.97. A strategy that lifts its win rate without handing the payoff back has changed something other than its exits.
mt5depot EA pages publish win rate beside the trade count and the data window, so you can read the figure against the sample that produced it. Doing that across the fourteen runs here exposes a limit of the table above: four listings sit between 78% and 80%, inside the band this page calls grid-and-range and marks as a warning — yet none of the four runs a grid, since two enter on an Ichimoku cross and two on RSI. Two more, at 58.10% and 31.27%, land outside every band the table offers. The bands sketch exit design; they do not identify it. Each listing does clear the payoff its own win rate demands — the ratio (1 − p) ÷ p — but that demand swings from 0.26 at the 79.66% listing to 2.20 at the 31.27% one, and the room above it runs thinnest where the count runs highest: roughly a fifth of headroom at the top win rate against roughly a third at the bottom.
Common mistakes
Comparing win rates across strategy types
A trend EA and a scalping EA are supposed to have different win rates — the figures describe the exit design, not the quality. Comparing them ranks the design, not the edge.
Ignoring the sample size
Over fifty trades a win rate can drift by ten points on chance alone. It becomes a planning number in the hundreds, not the dozens.
Assuming the backtest win rate survives live execution
Spread and slippage push marginal winners into the loss column, and that effect lands hardest on strategies whose average win is small — exactly the high-win-rate ones.
Reading a shift between windows as evidence of curve fitting
It usually means the second window was small. Comparing each listing's backtest win rate with the rate recomputed from the trades that closed after it went on sale, the median move across the thirteen runs here that have post-listing trades is 6.1 points — but one listing jumped from 55.90% to 86.67% on 15 post-listing trades, and another from 30.99% to 42.86% on 7. At those sizes the shift measures the sample, not the exit rules. Read the trade count in the second window before concluding anything about fitting.
Worrying about the breakeven convention before checking whether it bites
Counting breakeven trades as wins can move a published figure, and any seller should state the convention — but the effect is often nil. Across the 9,702 trades in the ledgers published on this site, exactly seven closed at zero. The listing carrying the most has three of them in 423 trades, which moves its win rate by 0.7 of a point. Ask for the convention, then open the ledger and see whether the answer could change anything.
Frequently asked questions
- What is a good win rate for an MT5 EA?
- There is no good win rate in isolation. Trend-following EAs typically win 35–55% of trades and remain profitable because their winners are several times their losers; scalping EAs typically win 60–75% because their winners are small. The figure only becomes a judgement once the average win and average loss are known.
- Can an EA with a 30% win rate be profitable?
- Yes, and many trend-following strategies are. At a 30% win rate a strategy needs its average win to exceed roughly 2.4 times its average loss to break even, which a wide-target trend design regularly achieves.
- Why is a very high win rate a warning sign?
- Because the easiest way to produce one is to stop closing losing trades. Martingale and grid EAs post win rates above 90% by holding and averaging losers, so the losses accumulate as unrealised drawdown instead of appearing in the count.
Related articles
EAs where this matters
Listings whose published record is judged on Win rate.
