MASTERMINDREADING ROOM阅览室
From the desk工作台随笔

Win Rate Is Overrated

A 35% win rate can be more profitable than a 70% win rate. Expectancy — not how often you win — determines whether a system makes money.

Published发布 2026-07-20

A trading system that wins 35% of the time can be more profitable than one that wins 70% of the time. Win rate is only half the equation — the other half is how much you win versus how much you lose on each trade. Expectancy, not win rate, is the number that determines whether a system makes money over time.

The expectancy formula

Expectancy measures the average amount won or lost per unit of risk across all trades. Expressed in R multiples — where 1R is the amount risked per trade — it is:

Expectancy = (Win% × Average Win in R) − (Loss% × Average Loss in R)

Win% + Loss% = 100%
A positive expectancy means the system makes money over a large sample; negative means it loses.

R multiples give a unit-neutral way to measure outcomes. If you risk $200 on a trade and the trade makes $600, the outcome is +3R. If you lose, it is −1R. Working in R separates sizing decisions from edge measurement — a useful property when reviewing trades across different position sizes.

35% win rate beats 70% win rate: the arithmetic

Two systems, 100 trades each

System A — 35% win rate, 3:1 average win to loss ratio
Win% = 35, Average Win = 3R, Average Loss = 1R
Expectancy = (0.35 × 3R) − (0.65 × 1R) = 1.05R − 0.65R = +0.40R per trade

Over 100 trades at 1% risk per trade on a $50,000 account ($500 per trade):
Average profit = 100 × $500 × 0.40 = $20,000

System B — 70% win rate, 0.5:2 average win to loss ratio
Win% = 70, Average Win = 0.5R, Average Loss = 2R
Expectancy = (0.70 × 0.5R) − (0.30 × 2R) = 0.35R − 0.60R = −0.25R per trade

Over 100 trades at the same 1% risk:
Average loss = 100 × $500 × 0.25 = $12,500

System A wins 35% of the time and makes money. System B wins 70% of the time and loses money. The difference is entirely in the ratio of average winner to average loser.

System B is psychologically seductive precisely because of its high win rate. Winning 7 out of 10 trades feels good. But if each winner returns half a unit while each loser costs two units, the account bleeds steadily over time. This pattern — high win rate, poor reward structure — is how many intuitive trading styles produce inconsistent results despite frequent winning trades.

The common trap: confusing win rate with edge

Common trap Traders who focus on win rate tend to cut winners early (to lock in the win) and hold losers long (to avoid booking the loss). Both behaviors directly undermine expectancy: they shrink average winners and expand average losers, moving toward a high-win-rate, negative-expectancy profile. The emotional pull toward high win rates actively works against profitability.

The discipline that positive-expectancy systems require is usually the opposite of what feels comfortable. Trend-following approaches with 35–40% win rates often outperform over time precisely because they allow winners to compound while cutting losers promptly. The discomfort of frequent small losses is structural, not a signal that the system is broken. A trading journal that tracks R multiples per setup is the only reliable way to measure whether your actual expectancy matches your intended system.

Breakeven win rate as a function of reward ratio

For any given average win-to-loss ratio (the R:R ratio), there is an exact win rate at which expectancy equals zero. Any win rate above that level produces positive expectancy; below it produces negative expectancy.

Breakeven win% = 100 / (1 + R:R)
Average R:R (win/loss) Minimum win rate needed
0.5 : 1 (small winners vs losers)66.7%
1 : 1 (equal winners and losers)50.0%
1.5 : 140.0%
2 : 133.3%
2.5 : 128.6%
3 : 125.0%
4 : 120.0%
5 : 116.7%

The table reveals the design space. A system with a 2:1 average reward-to-risk ratio is profitable at any win rate above 33.3%. A system with 0.5:1 requires winning more than two-thirds of the time to break even. Every trading approach implicitly sits somewhere in this space — the question is whether the trader knows where.

Use the risk-reward calculator to compute the R:R on any trade before entry, given a specific entry price, stop level, and target. It also shows the expectancy contribution of a single trade given your historical win rate at that R:R.

What win rate is actually useful for

Win rate is not useless — it matters in at least two ways. First, very low win rates (below roughly 25%) create long losing streaks that can be psychologically difficult to sustain even when the system has positive expectancy. The article on losing streaks shows that a 35% win rate over 100 trades carries a roughly 99.5% chance of hitting at least a five-loss streak. Position sizing must account for this even when the long-run edge is positive.

Second, win rate interacts with expectancy stability. A system with a 70% win rate and positive expectancy will show more consistent short-term results than a 30% win-rate system with the same expectancy, because the high-win-rate system has lower per-period variance. This is not a reason to prefer high win rate — it is a reason to understand how many trades your edge requires before the expectancy becomes reliable.

Where this breaks

Expectancy assumes the distribution of wins and losses is stationary — that the win rate and average R values from your sample will persist into the future. In real markets they will not. Win rates shift with market regime, volatility conditions, and correlation regimes. A system with measured positive expectancy in one environment may have negative expectancy in another. This is why the journal and periodic recalculation matter: the expectancy calculation is not a one-time finding, it is a running measurement.

The formula also assumes trades are independent, which overstates reliability. When markets trend strongly in one direction, consecutive trades in a trend-following system will be positively correlated — both winning or both losing together. Correlated trade outcomes produce more variance than the independent model assumes, meaning that even a positive-expectancy system can underperform its expected value for extended periods during regime changes.