Why a 90% Win Rate Is a Warning Sign, Not a Feature

The number that grabs your eye on an EA sales page is almost always the win rate. "92% wins." "88% winning trades." It feels obvious — if it wins nine times out of ten, it must be good. In practice, the opposite paradox holds: the EAs that lead with a high win rate are the ones most likely to empty an account in a single move.

The reason is simple. Win rate, on its own, tells you nothing about how much you win and how much you lose on average.

Why win rate alone is meaningless — the math

What decides whether you make money is not the win rate, it is the profit factor (gross profit ÷ gross loss). The two are related like this:

Which leads to the key point: the same profit factor can be produced by wildly different win rates.

Same profit factor. Opposite character:

• The trend type is "many small losses + a few large wins." Each loss is small.
• The high-win-rate type is "many small wins + the occasional huge loss." One loss erases five or six wins in an instant.

When a losing streak hits the high-win-rate system — and it always does — months of profit vanish overnight. Same profit factor, completely different blow-up risk.

What hides behind a high win rate

The most common way to achieve a 90% win rate is, in fact, simple: never realize a loss.

• Averaging down (nampin): add to a losing position as it moves against you. The average entry drops, so a small bounce lets you close "in profit." Win rate goes up. But if it doesn't bounce, the position snowballs until the margin runs out.
• Martingale: double the lot after every loss. One win recovers everything. Win rate goes up. But a long losing streak blows up exponentially.
• Grid: place orders at fixed intervals in both directions. Range-bound markets give a high win rate. But the moment a trend appears, one whole side is underwater.

What they share is a structure: they buy the win rate by not cutting losses. The win rate just defers the risk into the future. That clean, rising backtest equity curve may simply mean "the one big loss hasn't arrived yet."

So look at max drawdown and whether it has a stop

When you evaluate a high-win-rate EA, the win rate is not the number to read.

1. Does every trade have a hard stop-loss? If not, the open loss can grow without limit.
2. Is the max drawdown equity-based? Balance drawdown looks like a fraction of reality for an EA that holds losers (see drawdown management).
3. Year-by-year P&L. A high-win-rate EA lets the average hide the year it blew up.

If the win rate is above 80% and no stop-loss is stated, that is a sign to suspect averaging or a grid.

The case for a 34.5% win rate

Most of our core trend EAs win well under half their trades. One USDJPY trend follower, over 9.4 years of real-tick testing: win rate 34.5%, profit factor 1.50, maximum equity drawdown 10.3%.

It loses two times out of three. But every one of those losses is held small by a hard stop, and the occasional large trend more than covers them. Because the losses are designed to be small from the start, no single trade can end the account. That is what a correctly built breakout trend follower looks like.

A low win rate is not a defect — it is evidence that the risk is being disclosed honestly. We do not lead with win rate because we know that the higher the win rate looks, the more likely something is being hidden behind it.

Check it yourself

This "low win rate, hard stop, no grid" trend EA is free and open-source. Run the backtest and you can see — in the code — exactly how a 34.5% win rate still produces a rising equity curve.

Donchian Trend Engine (free EA)

The same discipline as real-tick-verified products: MEGAMAX DONCHIAN (USDJPY), BITCOIN COMET (BTCUSD) and more — all hard stop, no grid, no martingale, with full real-data verification published.

Summary

• Win rate alone tells you nothing; profit factor is what matters.
• Same profit factor, but a high win rate carries far higher blow-up risk (the occasional huge loss).
• Most 90% win rates are built by avoiding stops via averaging, martingale or grids.
• What to read: hard stop-loss, equity drawdown, year-by-year P&L.
• A low win rate is not a flaw — it is honestly disclosed risk.

Related: We tested loss filters on every EA · EA drawdown management · 43 popular gold EAs, mass real-data backtest