Win Rate — and Why It Is Not Enough
Winning more than you lose sounds like the right goal. In systematic trading, it rarely is
Win rate is one of the most intuitive metrics in trading.
It tells you what percentage of trades closed in profit. A win rate of 70% sounds compelling. A win rate of 40% sounds worrying. But these numbers, taken in isolation, say almost nothing useful about whether a trading model actually has a positive edge.
What win rate actually tells you
Win rate measures frequency, not quality. It counts how often the model wins, but says nothing about how much it wins or how much it loses when it does not.
A model can have a win rate of 75% and still be unprofitable — if the 25% of losing trades are large enough to wipe out the gains from the winners. Conversely, a model with a win rate of 40% can be consistently profitable if its average winning trade is significantly larger than its average losing trade.
Win rate only becomes meaningful once it is read alongside the average win size and average loss size. That is the basis of Expected Value.
Win rate and Expected Value
Expected Value captures what win rate alone cannot. It asks: on average, what does one trade return?
The simple form is: EV = (Win Rate × Average Win) − (Loss Rate × Average Loss)
A positive Expected Value means the model has a statistical edge. A negative one means it does not — regardless of how high the win rate is. Two models can have identical win rates but completely different Expected Values depending on how their winning and losing trades are sized.
This is why win rate is a poor shortcut for evaluating a trading model. High win rate combined with poor risk/reward is a common pattern among fragile strategies — they feel consistent until a small number of large losses erodes months of small gains.
Win rate and risk/reward
There is a direct relationship between win rate and risk/reward ratio that determines whether a model is viable.
A model targeting twice the reward of its risk requires only about one win in three to break even. A model with a 1:0.5 risk/reward needs to win two out of three just to stay flat. Neither scenario is inherently better — what matters is whether the combination produces a positive Expected Value consistently.
A model engineered to maximise win rate — by using wide take-profits or cutting winners short — often looks attractive until the average loss eventually dominates. The win rate was real; the edge was not.
How darwintIQ evaluates models beyond win rate
In darwintIQ, win rate is shown as one data point within a broader set of metrics. The platform evaluates trading models using multiple dimensions simultaneously: Expected Value, Profit Factor, Drawdown, Return Stability, and Fitness — each of which provides context that win rate alone cannot.
The Fitness score incorporates win rate as one factor among several. A model with a high win rate but poor drawdown behaviour, low return consistency, or an insufficient number of trades will not score well on Fitness regardless of that win rate figure. The platform is designed specifically to avoid the trap of optimising on a single superficially appealing metric.
This is particularly relevant in an adaptive system. A model that has been achieving a high win rate in a recent trending market may show that win rate collapse quickly when the regime shifts. Evaluating win rate in the context of full model quality — rather than as a standalone figure — is what makes the difference between temporary performance and structural reliability.
Final thoughts
Win rate is not useless. It is part of the picture. But it is a dangerous metric to rely on in isolation. The real question is always: what does each trade return on average, and how consistent is that return? Those are the questions that Expected Value, Profit Factor, and the broader model profile are built to answer.