A win rate above eighty percent should raise eyebrows
A system that wins 85 percent of the time sounds like a dream. In scalping, it is often a warning.
High win rates can come from a structure with small targets and large stops. The trader wins often but loses big when it goes wrong. The expectancy may be negative even if the hit rate is spectacular.
Another source is data. A backtest with unrealistic fills, ignoring spread, or ambiguous bar order can produce inflated win rates that vanish in live trading.
Martingale and grid methods also create high win rates, hiding the tail risk until a single sequence erases the account.
The right question is not how often the strategy wins, but how much it makes on average per trade after costs, how deep the worst drawdown has been, and how the results look under realistic conditions.
A lower win rate with a good reward to risk ratio is often sturdier than a high one that depends on rare large losses not occurring.
Another red flag is a curve that rises in a perfectly straight line. Real strategies have texture. A smooth line suggests either an averaging-down method or a curve fitted to the past.
What is the average win versus the average loss behind the favourite strategy's win rate?
A win rate above eighty percent should raise eyebrows
-
FXS HongKong Scalper
- Posts: 543
- Joined: Sat Oct 24, 2026 10:15 pm
Re: A win rate above eighty percent should raise eyebrows
The small target, large stop structure is the usual reason. Checking the average loss divided by average win is a quick way to see if a high win rate is genuine or a hidden risk.
FXS Official Team | Hong Kong Session Desk
-
Amanhasnoname
- Posts: 2060
- Joined: Fri Oct 30, 2026 9:45 am
Re: A win rate above eighty percent should raise eyebrows
I learned this one the expensive way with a bought robot a few years back. The sales page showed something like an eighty-eight percent hit rate on EURGBP, and for the first five weeks live it really did win almost every day. What I hadn't looked at closely was that the take profit sat at four pips while the stop was out past thirty. One quiet Thursday the pair slid through the stop on a single push, and that loss wiped roughly six weeks of small wins in about two minutes.
After that I started keeping a simple sheet for anything I test: average winner, average loser, and the ratio between them next to the win rate. My own scalping setup now wins a little under half the time, but the average win is close to double the average loss once spread and commission are taken out, and the equity curve is far less nerve-wracking because no single trade can undo a month.
The straight-line curve point is the one I'd underline. That robot's backtest looked like a ruler, and I should have taken that as the warning it was.
After that I started keeping a simple sheet for anything I test: average winner, average loser, and the ratio between them next to the win rate. My own scalping setup now wins a little under half the time, but the average win is close to double the average loss once spread and commission are taken out, and the equity curve is far less nerve-wracking because no single trade can undo a month.
The straight-line curve point is the one I'd underline. That robot's backtest looked like a ruler, and I should have taken that as the warning it was.
Re: A win rate above eighty percent should raise eyebrows
FXS HongKong Scalper wrote: Thu Nov 12, 2026 1:17 pmThe small target, large stop structure is the usual reason. Checking the average loss divided by average win is a quick way to see if a high win rate is genuine or a hidden risk.
Loss/win ratio is useful, but I would pair it with costs and the distribution of losing streaks. Suppose a system wins 82% with an average winner of 0.25R and an average loss of 1.4R: expectancy before costs is 0.82 × 0.25 − 0.18 × 1.4 = 0.205 − 0.252, or −0.047R per trade. A 90% win rate can have the same problem if the tail loss is large enough. Add round-trip spread and commission in R, then check maximum drawdown and the worst run, not just the mean. I would also compare filled trades with all valid signals; limit fills can be selected by adverse continuation, while unrealistic backtest fills can inflate the ratio. At 0.5% risk, twenty consecutive losses compound to about a 9.5% drawdown. A smooth curve with no such cluster may simply be a short sample. Win percentage only makes sense beside payoff ratio, costs, sample size, and tail risk.
It’s Fairman 