Trading Expectancy: The Formula Every Model Should Pass
Win rate alone tells you nothing. Risk/reward alone tells you nothing. Expectancy combines both into a single answer.
Trading expectancy answers the question every model design eventually arrives at: on average, how much money does this strategy make per trade?
It is the bridge between win rate and risk/reward. Neither metric alone is meaningful — a 70% win rate is worthless if the losses are three times the size of the wins, and a 1:5 risk/reward is worthless if the model only wins 10% of the time. Expectancy collapses both into a single number that tells you whether the underlying edge exists at all.
The expectancy formula
The standard formula:
Expectancy = (Win Rate × Average Win) − (Loss Rate × Average Loss)
Where loss rate is simply 1 minus the win rate, and the values are typically expressed in money or in R-multiples (multiples of the risk taken on the trade).
A worked example. A model wins 45% of its trades, with an average win of 1.6R and an average loss of 1.0R:
Expectancy = (0.45 × 1.6) − (0.55 × 1.0) = 0.72 − 0.55 = 0.17R per trade
That means, on average, the model produces 0.17 times its risk per trade taken. Over 200 trades, that compounds into 34R — assuming the inputs hold up.
Positive expectancy is necessary, not sufficient
A model with positive expectancy has, in the sample measured, an edge. That is the floor. It does not tell you:
- Whether the edge will survive in live conditions
- How smooth or volatile the equity curve will be
- How long the drawdowns last
- Whether the next 50 trades will look like the last 200
This is why darwintIQ's evaluation does not stop at average per-trade outcomes. The Stability Score, Return Stability, and distribution distance metrics together describe whether the expectancy is the product of a consistent pattern or a few lucky outliers.
Why expectancy can mislead
Two warnings worth keeping front of mind.
Outlier risk. A single very large win in the sample can pull the average win sharply upward and produce a deceptively healthy expectancy. The same number with the outlier removed might be near zero or negative. Robust evaluation requires looking at the distribution, not just the mean.
Sample size. Expectancy calculated on fewer than around 100 trades is statistically unreliable. The fewer trades, the more the figure is driven by chance. darwintIQ's rolling 4-hour evaluation window weighs this implicitly by combining the metric with measures of how the model's behaviour holds up over time.
How expectancy fits with other metrics
Expectancy answers "is there an edge per trade?". To turn that into a verdict on the model as a whole, it pairs naturally with:
- Profit Factor — the ratio of gross profit to gross loss, which captures the same idea from a different angle
- Drawdown — how painful the path to that expectancy will be
- Frequency of trades — high-expectancy models with very few trades produce capital inefficiency; low-expectancy models with many trades suffer from accumulated costs
On the dashboard, you will not see a single field labelled "Expectancy", but you can derive it from the win rate and risk/reward figures shown in the Trader Detail view. The dashboard's design assumes expectancy is the consequence of the deeper metrics, not the diagnostic itself.
Final thoughts
Expectancy is one of the few formulas in quantitative trading that almost everyone agrees on. The disagreements happen at the next layer: how stable that expectancy is, whether it will survive translation to a different market regime, and how much capital should be deployed against it. A model with strong expectancy and weak stability is more dangerous than one with modest expectancy and steady distribution behaviour. The math is the first step. The discipline is in not stopping there.