Two traders show you their last hundred trades. The first won 60 of them. The second won 40. Asked which is the better trader, most people answer in under a second, and they pick the first one.
The second trader made twice as much money.
Win rate is the most quoted number in trading and the most misleading one in isolation. It says how often you are right. It says nothing about how much you make when you're right or how much you lose when you're wrong — and those two figures decide everything. The number that combines all three is expectancy, and once you can read it, the win-rate argument stops being an argument. This is the formula, the breakeven table that comes out of it, the way a 90% win rate can still lose money, and how to compute your own honestly from trades you've already taken.
Key takeaways
- Expectancy = (win% × average win) − (loss% × average loss). A 40% win rate at 2.5:1 returns +0.40R per trade; a 60% win rate at 1:1 returns +0.20R. The lower win rate makes double.
- Every reward-to-risk ratio has a breakeven win rate: 1:1 needs 50%, 2:1 needs 33.3%, 3:1 needs 25%. Below that line the strategy loses money no matter how clean the entries look.
- A 90% win rate can still bleed. Winners of +0.3R against losers of −5R is −0.23R per trade — a losing strategy that is right nine times out of ten.
- Costs come straight out of expectancy. A +0.30R gross edge carrying 0.20R of fees, funding, and slippage is a +0.10R strategy — two-thirds of the edge gone before direction is involved.
- Two strategies with identical expectancy don't feel identical. The lower win rate waits 2.5 trades between wins on average and draws down deeper — same money, more discomfort, and that gap is where traders quit the better system.
The formula that ends the win-rate argument
Expectancy is the average amount you can expect to win or lose per trade, across a large number of trades, stated in the same unit every time:
expectancy = (win% × average win) − (loss% × average loss)Measure it in R — multiples of the amount you risked on the trade — and the position size drops out, so trades of different sizes become comparable. One R is your initial risk. A winner that made two-and-a-half times your risk is +2.5R. A full stop-out is −1R.
Now the two traders from the opening, with the missing numbers filled in. Both risk 1R per trade.
| Win rate | Average win | Average loss | Expectancy | Per 100 trades | |
|---|---|---|---|---|---|
| Trader A | 60% | +1.0R | −1.0R | +0.20R | +20R |
| Trader B | 40% | +2.5R | −1.0R | +0.40R | +40R |
Trader A: 0.60 × 1.0 − 0.40 × 1.0 = +0.20R. Trader B: 0.40 × 2.5 − 0.60 × 1.0 = +0.40R. In dollars, if 1R is $100, A clears $2,000 over the hundred trades and B clears $4,000. The trader who is wrong more often than right takes home double, because the size of B's wins does more work than the frequency of A's.
This is the whole point, and it is worth sitting with: win rate is one of three inputs, and on its own it cannot tell you whether a strategy makes money. A trader optimizing purely for a higher win rate — taking profit early to lock the green, giving losers room to come back — is often walking expectancy in the wrong direction while the scoreboard they're watching goes up.
The breakeven win rate for every reward-to-risk ratio
Set expectancy to zero and solve for the win rate, and you get the exact percentage you must clear to stop losing money at a given reward-to-risk ratio:
breakeven win rate = 1 ÷ (1 + reward-to-risk)| Reward-to-risk | Breakeven win rate | Read it as |
|---|---|---|
| 0.5:1 | 66.7% | Cutting winners short: you must be right two times in three just to tread water |
| 1:1 | 50.0% | The coin-flip line most traders picture by default |
| 1.5:1 | 40.0% | Being wrong 60% of the time is still profitable here |
| 2:1 | 33.3% | One winner covers two losers; a third of your trades can carry you |
| 3:1 | 25.0% | A 1-in-4 hit rate makes money if you let winners run |
| 4:1 | 20.0% | Eight losses in ten and still ahead — if the two winners are large |
The table is a reality check in both directions. If your setup pays 2:1 and you win 45% of the time, you are comfortably profitable and any instinct to "tighten up the win rate" is expectancy-destroying. If your setup pays 0.5:1 because you scalp small targets, a 60% win rate that feels good is below water — you need 66.7% just to break even, and few discretionary traders hold that line after costs.
The number that actually needs defending is not the win rate. It's the average loss. Every entry in this table assumes losses average exactly 1R. The moment a few losers run to 2R or 3R because a stop got widened, the breakeven win rate climbs and the whole column moves against you.
The 90%-win-rate account that still bleeds
The clearest way to feel how little win rate means alone is to break it on purpose. Here is a strategy that wins 90% of its trades:
| Win rate | Average win | Average loss | Expectancy | Per 100 trades |
|---|---|---|---|---|
| 90% | +0.3R | −5.0R | −0.23R | −23R |
0.90 × 0.3 − 0.10 × 5.0 = −0.23R. Nine trades in ten close green, the equity curve climbs in a satisfying staircase, and the account loses 23R over a hundred trades. This isn't a contrived edge case — it is the exact shape of the most common way retail accounts die: take profit fast for the dopamine of a win, hold the loser and "average in" because closing it would admit the miss, and let one trade in ten erase the other nine. A 90% win rate is not a good strategy. It is frequently the fingerprint of a bad one.
The mirror image is the trend trader who wins 35% of the time, sits through long strings of small losses, and makes a comfortable living on the occasional 5R runner. Their win rate would embarrass them at a party. Their expectancy pays the rent. This is also why a losing streak is weak evidence about a strategy — a genuinely profitable low-win-rate system spends most of its life in a drawdown that a high-win-rate trader would never tolerate.
Expectancy grid: find roughly where you stand
You don't need software to place yourself. Estimate your win rate and your typical reward-to-risk, then read the cell. Every value is expectancy in R per trade, assuming losses average 1R.
| Win rate ↓ R:R → | 1:1 | 1.5:1 | 2:1 | 3:1 |
|---|---|---|---|---|
| 30% | −0.40R | −0.25R | −0.10R | +0.20R |
| 40% | −0.20R | 0.00R | +0.20R | +0.60R |
| 50% | 0.00R | +0.25R | +0.50R | +1.00R |
| 60% | +0.20R | +0.50R | +0.80R | +1.40R |
| 70% | +0.40R | +0.75R | +1.10R | +1.80R |
Two things fall out of the grid. First, the profitable region is large — there are many honest ways to make money, and they don't agree on win rate at all. A 30% win rate at 3:1 (+0.20R) and a 60% win rate at 1:1 (+0.20R) are the same business. Second, the way you move through the grid matters as much as where you sit. Pushing a 50%/2:1 setup (+0.50R) toward a higher win rate by exiting at 1:1 drops you to +0.20R. You'd feel more successful and earn less than half as much.
Costs come out of expectancy, not out of nowhere
Every figure so far is gross. The market does not hand you gross. Fees, funding payments on perpetual positions, and slippage on entry and exit are all subtracted from the same expectancy number — and because expectancy is often a small figure, small-looking costs are not small.
| R per trade | |
|---|---|
| Gross expectancy | +0.30R |
| Trading fees | −0.03R |
| Funding paid on held positions | −0.10R |
| Slippage vs planned entry/exit | −0.07R |
| Net expectancy | +0.10R |
Two-thirds of the edge disappeared before a single directional call was involved. Over 100 trades that is the difference between +30R and +10R — the same strategy, three times the take, decided entirely by leakage the trader never itemized. Funding especially hides, because it accrues quietly while a position is open rather than showing up as a line on the fill. The practical rule: your real expectancy is the gross number minus every cost, and until you've subtracted them you don't know your edge — you know your gross. Cost data needs no large sample to trust; it is arithmetic on trades that already happened, and it is usually the fastest 0.1R a trader can recover.
"What's my expectancy per setup, net of fees and funding — and which one is quietly below breakeven?"
That answer lives in your own trade history and nowhere else. A coach grounded in your journal can split expectancy by setup, subtract the costs attached to each trade, and tell you that the setup you trade most is the one dragging the book. A general chatbot can only re-explain the formula, because your fills are the one input it doesn't have. The version worth acting on is computed from your trades — and it frequently disagrees with the setup you're most attached to.
Why the higher-expectancy strategy can feel worse
Expectancy tells you what a strategy earns. It says nothing about the ride, and the ride is where decisions actually get made. Take two strategies with identical expectancy of +0.20R:
| Win rate | R:R | Expectancy | Avg trades between wins | |
|---|---|---|---|---|
| Smooth | 60% | 1:1 | +0.20R | 1.7 |
| Lumpy | 40% | 2:1 | +0.20R | 2.5 |
Same money. Very different experience. The lumpy strategy waits longer between wins, strings more losses together, and puts you underwater more often even though it ends in the same place. Most traders can't sit in the lumpy one — they abandon it during a normal losing cluster, right before the large winner that makes the average true. This is the quiet reason win rate seduces people: a higher win rate buys a smoother equity curve, and a smoother curve is easier to hold. That comfort is worth something real. It is just not worth confusing with profit, and it is not worth buying by exiting winners early, which pays for the smoothness with your edge.
Note too that any single number here is an estimate with error bars. An expectancy computed from 30 trades barely constrains the truth; a modest edge takes a couple hundred trades to separate from zero with confidence. Expectancy is the right target. It is also slow to measure, which is one more reason to fix the costs first.
Five ways expectancy will mislead you
The metric is load-bearing, and like any single number it can be over-trusted. Each of these has a cost.
- A small sample fakes a big edge. Twenty trades at +0.8R is not an +0.8R strategy; it is a wide interval that happens to be centered high today. Compute expectancy, but read it alongside how many trades are behind it.
- Averages hide the trade that matters. One 15R outlier can carry an entire average and will not repeat on schedule. Look at the median trade and the result with your single best trade removed — if the edge lives entirely in one fill, you don't have an edge yet, you have a story.
- Expectancy per trade ignores time. +0.30R in two hours and +0.30R over three days are not the same business. Rank setups by R per day, not R per trade, when capital is the constraint.
- Gross expectancy is not your expectancy. Until fees, funding, and slippage are subtracted per trade, the number is a fiction that flatters you. This is the most common way a "profitable" journal is actually flat.
- A past average is not a promise. Regimes change; a setup with a real +0.4R edge last quarter can be flat this one. Expectancy describes the trades you took, under the conditions they happened in — it is evidence, not a guarantee.
Where to start
Pull your last 50 to 100 closed trades and compute one number: expectancy in R, net of costs. Win rate times average win, minus loss rate times average loss, with fees and funding already subtracted from each result. It takes about twenty minutes in a spreadsheet, and it is the single most clarifying figure most traders have never calculated for themselves.
Then split it by setup, and expect a surprise. The setup you trade most is often not your best one, and the one you dismiss may be quietly carrying the book. Once you can see expectancy per setup, the sequence of decisions gets simple: cut or fix anything below breakeven after costs, size the highest-expectancy setups the largest, and stop optimizing for a win rate that was never the thing paying you. Sizing is the direct lever on expectancy — how much you risk per trade multiplies whatever edge the number reveals, in both directions.