Win rate doesn't tell you if a trading strategy works. Trading expectancy does. Here is how to calculate it and what it actually measures.
Trading expectancy is the expected profit or loss per trade, expressed as a single number that tells you whether a strategy makes money over many trades. It combines win rate, average win, and average loss into a complete picture. Win rate alone tells you nothing about whether a strategy is profitable. Expectancy tells you everything that win rate cannot.
Most traders evaluate strategies primarily on win rate. A strategy winning 65% of trades sounds better than one winning 45%. This is only true if the size of wins and losses is identical. In practice, it almost never is.
A strategy with a 65% win rate but average losses 3x the size of average wins is a losing strategy. Every 100 trades: 65 wins of $100 = $6,500. 35 losses of $300 = $10,500. Net loss: $4,000. The 65% win rate masked a fundamentally unprofitable expected value.
A strategy with a 40% win rate but average wins 4x the size of average losses is a strong strategy. Every 100 trades: 40 wins of $400 = $16,000. 60 losses of $100 = $6,000. Net profit: $10,000. The 40% win rate sounds poor. The expectancy is excellent.
This is why systematic traders do not optimize for win rate. Win rate is one input into expectancy, not the final measure. The output is the expected value per trade — the number that determines whether the strategy produces profit over time regardless of how individual trades feel in the moment.
The formula: Expectancy = (Win Rate x Average Win) minus (Loss Rate x Average Loss). Loss Rate equals 1 minus Win Rate.
Example 1 — Positive expectancy: Win rate 55%, average win $150, average loss $100. Expectancy = (0.55 x $150) minus (0.45 x $100) = $82.50 minus $45.00 = $37.50 per trade. Over many trades, this strategy expects to produce $37.50 per trade on average.
Example 2 — Negative expectancy despite high win rate: Win rate 70%, average win $80, average loss $200. Expectancy = (0.70 x $80) minus (0.30 x $200) = $56 minus $60 = negative $4 per trade. Loses money over time despite winning 70% of trades.
Expectancy can also be expressed as an R-multiple: expectancy per dollar risked. If average risk per trade is $100 and expectancy is $37.50, the R-multiple is 0.375R. This normalizes expectancy across strategies with different absolute position sizes, making it easier to compare strategies directly.
To calculate expectancy from a backtest, collect the dollar value of every win and every loss over the full sample period. Calculate the average win, average loss, and win rate. Apply the formula. The result is the average expected outcome per trade if the same conditions repeat.
Zero expectancy means the strategy breaks even before costs. After spread, fees, and slippage, it is a losing strategy. Zero expectancy is not a viable minimum — it requires perfect, cost-free execution that does not exist in live markets.
Positive expectancy below 0.25R is marginal edge. The strategy makes money but provides little buffer above zero. Small execution inefficiencies, slightly wider spreads, or changes in market conditions may erode or eliminate the edge. Requires careful ongoing monitoring to confirm it persists.
Positive expectancy between 0.25R and 0.50R is solid edge. This is the range where most well-designed systematic strategies operate. Enough margin above zero to survive normal variation in market conditions and execution quality.
Positive expectancy above 0.50R is strong edge. Unusual in liquid, efficient markets. When a backtest produces expectancy significantly above 0.50R, examine the methodology carefully for overfitting, lookahead bias, or survivorship bias before accepting the result as genuine.
Negative expectancy cannot be fixed by position sizing, discipline, or risk management. No management technique converts a strategy with negative expected value into a profitable one over sufficient sample size. The only fix is changing the underlying strategy.
Aggregate expectancy conceals regime-specific performance in the same way aggregate win rate does. A strategy with positive overall expectancy of 0.30R might have expectancy of 0.65R in trending conditions and expectancy of negative 0.15R in ranging conditions. The combined figure looks acceptable. The regime-segmented analysis reveals two different strategies: one with genuine edge and one that is eroding it.
Separating expectancy by market regime is the most important diagnostic step after the basic calculation. It reveals whether the strategy has genuine edge in its intended regime, whether the regime filter is working as designed, and whether there is an unexploited opportunity in the opposite regime.
A trend-following strategy should have positive expectancy in trending conditions. If it does not, the signal logic is wrong regardless of what the regime filter is doing. If the regime filter is working correctly, the combined expectancy (after filtering) should be higher than the unfiltered expectancy because losing trades in the wrong regime have been removed.
If ranging-condition expectancy is strongly negative, the regime filter is the key improvement. If it is near zero or slightly positive, there may be a mean-reversion component worth adding. The numbers from expectancy segmentation tell you what to build next.
In the live signal pipeline, expectancy is calculated separately for each regime state, each confidence band, and each pair. This produces a grid of expectancy values rather than a single aggregate number.
The most diagnostic finding from this segmentation: expectancy in the 90 to 96% confidence band for RANGING regimes was negative across the first several months of operation. High-confidence ranging signals were losing trades. This was counterintuitive — higher confidence should produce better outcomes. Investigation revealed that high confidence in a ranging regime correlated with specific indicator combinations that were technically confirming but reflected the late stages of a ranging period approaching a trend transition. These signals were high-confidence ranging entries at the worst possible moment in the regime cycle.
Segmenting expectancy by confidence band, not just by regime, was the diagnostic that revealed this pattern. Aggregate RANGING expectancy had masked it. The solution was not to improve confidence scoring but to add a trend-transition detection check before executing high-confidence ranging signals. A single aggregate expectancy number never reveals a pattern like this — it requires granular segmentation.
For the backtesting methodology that produces reliable expectancy data, see How to Backtest a Trading Strategy.
Small sample size. Expectancy from 20 trades is statistically unreliable. The observed win rate can deviate from the true win rate by 20 percentage points or more in a 20-trade sample purely by chance. At minimum 100 trades are required before expectancy provides meaningful signal. Several hundred trades across multiple market phases is the standard for robust expectancy estimation.
Non-stationarity. Markets change. The expectancy measured on 2021 data may not hold in 2024 data because the market's volatility regime, liquidity, or participant behavior has changed. Historical expectancy is a snapshot of past conditions. Ongoing live tracking with shadow data is the only way to monitor whether historical expectancy persists in current conditions.
Outlier sensitivity. If average win is driven by a small number of very large wins, positive expectancy may not be robust. A 100-trade sample with 39 standard wins of $50, one outlier win of $5,000, and 60 losses of $100 shows positive expectancy — but remove the outlier and expectancy turns negative. Outlier-driven expectancy is fragile. Check whether expectancy remains positive when the largest 5% of wins are excluded from the calculation.
Costs not included. Expectancy calculated before trading costs is not the expectancy experienced in live trading. Spread, fees, and slippage reduce expectancy directly. For strategies with high trade frequency or those using market orders on wide spreads, costs can reduce or eliminate a positive-expectancy edge entirely. Always include realistic cost estimates in the calculation before treating results as actionable.
Trading expectancy is the expected profit or loss per trade, calculated as: (win rate x average win) minus (loss rate x average loss). A positive result means the strategy is expected to produce a profit over many trades. A negative result means it is expected to produce a loss regardless of win rate, position sizing, or trade management. Expectancy is the single most informative metric for evaluating whether a trading strategy has genuine edge.
Collect all trade results from a backtest or live trading history. Calculate the win rate (number of winning trades divided by total trades), the average winning trade value, and the average losing trade value. Apply the formula: (win rate x average win) minus (loss rate x average loss), where loss rate equals 1 minus win rate. The result is the expected value per trade. Divide by average risk per trade to express as an R-multiple.
Positive expectancy above 0.25R (0.25 times the average risk per trade) is the minimum for a strategy worth trading after accounting for execution costs. Expectancy between 0.25R and 0.50R represents solid edge for a systematic strategy. Above 0.50R is strong and should be examined for overfitting before being accepted. Zero expectancy, while technically break-even, becomes negative after spread, fees, and slippage. Negative expectancy cannot be managed into profitability — it requires changing the strategy.
Expectancy is more important. Win rate is one input into expectancy. A 70% win rate with large losses relative to wins produces negative expectancy and loses money over time. A 40% win rate with wins 3x the size of losses produces strong positive expectancy and makes money over time. Win rate feels important because humans track it naturally. Expectancy is the number that actually determines whether a strategy is profitable, independent of how wins and losses feel individually.
Positive expectancy means a trading strategy is expected to produce a net profit over many trades. It occurs when the combination of win rate and average win size outweighs the combination of loss rate and average loss size. A strategy does not need a high win rate to have positive expectancy — it needs either a sufficiently high win rate, sufficiently large wins relative to losses, or both. Positive expectancy before trading costs is necessary but not sufficient — it must remain positive after realistic cost estimates are included.
The Kelly Criterion uses expectancy to calculate the optimal position size as a fraction of account equity. The Kelly formula requires the win rate and the win/loss ratio, which are the components of expectancy. A higher expectancy (for a given win/loss ratio) produces a higher Kelly fraction, meaning larger optimal position sizes. In practice, most systematic traders use a fraction of the full Kelly (half or quarter Kelly) to reduce variance while maintaining positive expected growth. Kelly and expectancy are related frameworks for thinking about how much to bet given a known edge.
An R-multiple expresses trade outcomes as multiples of the initial risk (R) on the trade, where R is the distance from entry to stop loss. A trade that wins 00 on a 00 stop is a 3R win. A trade that loses the full stop is a 1R loss. Expressing results in R-multiples normalizes performance across different position sizes and markets. Expectancy expressed in R-multiples — the average R across all trades — tells you how much you expect to make per unit of risk, which is more meaningful than dollar amounts alone.