Win rate is a clue. Not the full equation.
A strategy can win most trades and still lose money—or win fewer than half and show positive mathematical expectancy. The missing ingredient is the size of its wins relative to its losses, after costs.
Updated 14 Nov 2026 · Original educational material · Hypothetical examples, not live market data
What is trading expectancy?
Expectancy is the model's average result per trade across a large set of hypothetical repeats. If we measure every trade against one unit of planned risk (1R), a full loser costs −1R and a 2R winner pays +2R. The expected gross outcome is the win probability multiplied by the winner payout, minus the loss probability multiplied by the loss size.
For example, an assumed 40% win rate with 2R winners produces 0.40 × 2R − 0.60 × 1R = +0.20R per trade. This is a conditional arithmetic average, not what any given trade earns. Actual winners or stop losses often vary, and an estimated past win rate may not hold in the future.
Why 50% is not a universal break-even point.
With wins and losses the same size (1R each), you need a 50% win rate to break even before trading costs. If winners pay 2R and losers cost 1R, the gross break-even win rate falls to 33.3%. Conversely, tiny average winners and large losses can make even a high win rate unprofitable.
The same win rate can mean different outcomes.
| Assumed win rate | Average winner | Gross expectancy | Interpretation |
|---|---|---|---|
| 70% | 0.3R | −0.09R | Negative despite many winners |
| 40% | 1.5R | 0.00R | Gross break-even |
| 40% | 2R | +0.20R | Positive gross expectation |
| 35% | 2R | +0.05R | Small gross edge, before costs |
Notice that the first row wins seven out of every ten trades in the assumed distribution yet has negative expectancy: 0.70 × 0.3 − 0.30 × 1 = −0.09R. The rows do not assert that any real strategy has these properties.
A tiny edge may disappear after fees.
Assume a 35% chance of earning +2R and a 65% chance of losing −1R. Gross arithmetic expectancy is +0.05R per trade. If all trading costs averaged 0.08R per trade under the same assumptions, the net expectancy would be −0.03R per trade.
Costs in R units depend on the actual risk amount. A round-trip fee of $8 is 0.08R when 1R equals $100, but only 0.016R when 1R equals $500. Funding, spread, partial exits and adverse fills further complicate comparisons.
Positive expectancy is not a smooth equity curve.
Even when the assumed average is positive, the order of wins and losses affects cash drawdowns. Ten losses in a row can occur in a model with a profitable average. Risking a fixed fraction of current capital also creates compounding effects: arithmetic expectancy in R does not directly equal long-run geometric account growth.
If every trade risks a fraction f of current equity and every win pays R, an idealized win multiplies equity by (1 + fR), while a loss multiplies it by (1 − f). Large swings can reduce geometric growth even where simple average return is positive. Explore the Risk of Ruin and Monte Carlo models for those distinct questions.
Three common misunderstandings.
“I win 80%, so I am profitable.”
Not necessarily. A few losses can outweigh many tiny wins. Compare the realized sizes of both sides of the distribution after trading costs.
“I target 3R, so I earn 3R on winners.”
An intended target is not a realized average. Partial closes, trailing stops and missed targets change what the strategy actually earns.
“Positive backtest expectancy guarantees profits.”
Sample size, changing markets, data leakage, selection bias and execution differences can make a historical estimate unreliable. A model's win probability is not known merely because you entered it.
Frequently asked questions.
Does a higher reward/risk always improve the strategy?
No. Larger targets may be reached less often. Changing a target can alter the win rate, holding period, fees and actual payoff distribution. Evaluate all of them together.
Is a 40% win rate good or bad?
By itself, neither. With constant 1.5R wins and 1R losses it is gross break-even; with 2R wins it is positive before costs. Real outcome sizes may differ.
Is win rate the same as probability of success?
An observed historical win fraction is an estimate from a particular sample and rule set. It does not prove the probability of the next trade or a stationary future success rate.
What if my losses are not always 1R?
Use the more general expectancy equation E[R] = p × average winning R − (1−p) × average losing R, then subtract costs. If outcomes have more than two categories, use their weighted average instead.