What is an R-multiple?
One R is the amount you risked on a trade — the distance from entry to stop, multiplied by position size. A trade that made three times what it risked is a +3R; one stopped out is a −1R.
The point of the unit is that it normalises everything. A gold trade risking $200 and a EUR/USD trade risking $50 are not comparable in dollars, and are directly comparable in R. So are results from an account that grew fivefold over the period being measured.
It also makes the arithmetic of a strategy legible. If your average win is 2R and your average loss is 1R, you need a win rate above roughly 33% to break even before costs — and that relationship is visible immediately, whereas in dollars it takes a spreadsheet.
The requirement is a defined risk per trade. A strategy without a stop has no R, which is one of several reasons discretionary traders find their results hard to evaluate.
How do you calculate expectancy?
Expectancy is the average R across every trade, wins and losses together — the sum of all R-multiples divided by the number of trades.
The equivalent form, which is more intuitive, is (win rate × average win in R) − (loss rate × average loss in R). A strategy winning 40% of the time with 2.5R winners and 1R losers has an expectancy of (0.4 × 2.5) − (0.6 × 1) = +0.4R.
That means each trade is worth 0.4 times whatever you risked, on average, over a large enough sample. Risk 1% per trade and the strategy earns roughly 0.4% per trade in expectation — which makes the frequency question immediately concrete.
Costs belong inside the calculation, not beside it. Subtract spread, commission and realistic slippage from every trade before computing R, because a strategy with +0.15R gross and 0.2R of costs has a negative expectancy and a perfectly healthy-looking win rate.
Why does expectancy beat win rate?
Win rate answers how often you are right, and being right is not what pays — a strategy can be right 80% of the time and lose money if the 20% are large enough.
The clearest illustration is the two-sided one. A rule with a 75% win rate, 1R winners and 4R losers has an expectancy of (0.75 × 1) − (0.25 × 4) = −0.25R and loses steadily. A rule with a 30% win rate, 4R winners and 1R losers has an expectancy of +0.5R and makes money. The second one feels much worse to trade.
This is why high win rates are the most reliably marketed number in trading and the least informative. Any strategy can be given a high win rate by widening the stop and tightening the target; the trade-off simply moves to the size of the losses, where it is less visible.
Expectancy makes the trade-off explicit, which is uncomfortable and useful in the same measure.
How many trades before expectancy means anything?
For the small edges most retail strategies actually have, hundreds of trades rather than dozens — and the weaker the edge, the more trades it takes to distinguish from noise.
The intuition is that expectancy is an average, and the reliability of an average depends on the spread of what is being averaged. R-multiples have a wide spread by construction: a trade set containing −1R losses and +6R winners has a standard deviation several times its mean, so the mean converges slowly.
A rough rule that gets the order of magnitude right: to distinguish an expectancy of +0.2R from zero, you need enough trades that the standard error is well under 0.2R, and for typical R distributions that lands in the several-hundred range.
This is the arithmetic behind the advice that a promising forty-trade result is not a result. It is not scepticism about the strategy; it is a statement about what forty samples of a wide distribution can support.
How do you track expectancy in live trading?
Record the intended risk on every trade at the moment of entry, before the outcome is known, and compute R from that rather than from what actually happened.
That ordering is the whole discipline. Deriving R after the fact — from where the stop ended up after being moved, or from the largest adverse excursion — produces a number that quietly incorporates hindsight and will not match the backtest it is being compared with.
Track partial exits explicitly. A trade scaled out at 1R and 3R has a blended R that needs computing rather than picking, and the two common shortcuts (taking the final exit, or the best one) bias in opposite directions.
Then compare the live expectancy against the backtested one on the same terms — same cost assumptions, same definition of R. A gap between them is the single most useful diagnostic a systematic trader has, because it localises the problem to execution rather than to the idea.