Net Pips vs. Win Rate — Which One Actually Measures Edge

Win rate is the first statistic almost everyone reaches for when judging a trading model, and it is the one most likely to mislead them. A high percentage of winning trades feels like direct evidence that a strategy works. But win rate says nothing about how large the winners are relative to the losers — and without that, it cannot tell you whether the model makes money. The figures that actually measure edge are net pips and per-trade expectancy.

The Statistic That Hides the Answer

Consider a model that wins 70% of its trades. That sounds excellent. Now suppose each winner gains 5 pips and each loser costs 15 pips. Across 100 trades, the 70 winners produce 350 pips and the 30 losers cost 450 pips, for a net of minus 100 pips. The model wins far more often than it loses and is steadily unprofitable.

The reverse case is just as common. A model that wins only 40% of its trades but gains 25 pips on winners while losing 8 pips on losers nets 1,000 pips against 480, a clear positive. Its win rate would look unimpressive on a leaderboard, yet it is the better model. Win rate, on its own, ranks these two backwards.

Net Pips: The Bottom Line

Net pips is simply the sum of every winning trade's gain minus every losing trade's loss over the evaluation window. It is the bottom line, and it cannot be gamed by trade selection the way win rate can — there is no way to make net pips look good without actually having made more than you lost. This is why it appears directly on the Out-of-Sample Holdout card in the Trader Detail View: when the question is whether a model stayed profitable on unseen data, net pips answers it without ambiguity.

Net pips does have a limitation: it scales with how many trades occurred. A model with 1,000 net pips over 200 trades and one with 1,000 net pips over 20 trades have produced the same total very differently. That is where expectancy comes in.

Expectancy: Edge Per Trade

Expectancy is net result divided by number of trades — the average pips a model produces each time it acts. It strips away how busy the model was and isolates the quality of each decision. A positive expectancy means that, on average, taking the trade is better than not taking it; a negative expectancy means the opposite, no matter how high the win rate.

Expectancy is the figure that makes win rate and average win/loss legible together. It encodes both the frequency of winning and the size of wins versus losses into one per-trade number, which is exactly why darwintIQ uses expectancy as the largest single component of its fitness function and reports per-trade expectancy on the holdout slice. If you only had room to look at one number, expectancy would be the one to choose.

Where Win Rate Still Belongs

None of this makes win rate useless — it makes it conditional. Win rate is meaningful once you read it alongside the average risk/reward, because the two only mean something in combination. A high win rate paired with a healthy average winner is genuinely strong. A high win rate paired with small winners and large losers is the trap above. Win rate also matters psychologically and operationally: a profitable model with a 35% win rate puts a trader through long losing streaks that a 60% model would not, even if both are net positive.

Used this way, win rate is one input. It describes the shape of the edge — how it is distributed across trades — rather than whether the edge exists. The existence question belongs to net pips and expectancy.

Final Thoughts

The most reliable way to misjudge a trading model is to rank it by win rate alone. Net pips tells you whether the model made money over the window; expectancy tells you how much edge sits in each individual trade; win rate, read together with average risk/reward, tells you the texture of how that edge arrives. Look at the bottom line and the per-trade edge first, and let win rate add detail to a picture the other two have already framed.