The average amount expected to win or lose per trade, calculated as win rate times average win minus loss rate times average loss. Positive expectancy is required for a profitable strategy.
also: mathematical expectancy, expected value per trade, edge
Updated
What one trade is worth on average, once the wins and the losses are weighed by how often each happens. If the figure is positive the strategy has an edge; if it is negative, nothing done with position size can turn it into a profit.
Expectancy is the only performance figure that answers the question everything else is a proxy for: does this strategy make money per trade, on average, after costs? Win rate and risk-reward are each half of it, and both are routinely quoted without the other.
Expectancy = (Win rate × Average win) − (Loss rate × Average loss) Loss rate is 1 − win rate, and average loss is taken as a positive number. Expressing it in R multiples instead of currency — average win and loss measured in units of the risk taken — makes it comparable across account sizes and lot settings.
No single number describes a strategy. These metrics change how this one should be read.
A strategy wins half its trades. Its winners average $120 and its losers average $80. Weighing each side by how often it happens gives the value of one trade.
The same $20 arrives as a long sequence of $120 gains and $80 losses, never as $20 payments — expectancy is an average, not a schedule.
Calculation (0.50 × 120) − (0.50 × 80) = 60 − 40 = $20
Result +$20 expected value per trade
Expectancy is a forecast built from three estimated inputs. Its reliability is the reliability of the weakest of them, so check where each number came from.
| Range | What it means |
|---|---|
| Computed from a backtest without realistic costs | Overstated. Spread and commission come straight out of the average win and straight onto the average loss, and the effect is largest where the targets are smallest. |
| Computed in currency on a fixed lot | Ulanternfish, but not comparable with another strategy tested at a different size. Convert to R multiples first. |
| Computed over several hundred trades | Reliable enough to plan around, because the average win is the input that needs the most trades to settle. |
| Re-computed on out-of-sample data | The figure worth acting on. A drop of a third from in-sample is normal; a sign change means the edge was fitted. |
| Recovered from a published summary | Four figures are enough: net profit, trade count, win rate and profit factor give back the average win and the average loss exactly. You do not need the seller to publish the three inputs. |
The fourteen runs published here work out to 1.84 USD through 20.52 USD of expectancy per trade, all at 0.1 lot on a 10,000 USD deposit. Read them as pre-commission figures: the manifests record commission 0 for all fourteen, so only spread and swap are inside those averages. Twelve of the fourteen also run on ticks the tester generates from M1 bars rather than on real ticks, and the spread those bars carry was measured on exactly one run, at 19.14 points.
Sizing changes the variance and the speed, never the sign. A strategy that loses a small amount per trade on average loses a proportionally larger amount when traded larger; martingale sizing simply concentrates the same negative expectation into rarer, bigger losses.
It is a long-run average over a distribution. At a 40% win rate the modal outcome of the next trade is a loss, and the account can sit below its starting point for dozens of trades while the expectancy is unchanged.
A backtest expectancy of $20 per trade routinely becomes $12 live once spread, commission and slippage are applied. The gap is not a surprise — it is the part of the calculation that was left out.
An expectancy of 0.2R and one of $20 cannot be compared until the risk per trade behind the second is known. Pick one basis and state it.
R needs a stated risk per trade, and listings frequently do not state one. Of the fourteen runs published here, two expose no stop-loss input at all, three set the stop as a percentage of price and nine set it in pips — so the risk behind each trade is not even in a common unit across the catalogue. The currency expectancy is recoverable for all fourteen; the R version is not.
The two trade off against each other. Among the fourteen runs here the win/loss size ratio runs from 0.33 to 2.92, and it moves opposite the win rate: the 79.66% winner takes losses three times the size of its wins, while the 31.27% winner takes wins nearly three times the size of its losses. Expectancy is what survives that trade-off; win rate on its own is not.
Listings whose published record is judged on Expectancy.
