# What Is Expectancy in Trading? Why Win Rate Misleads
Source: https://theoptionsbench.com/what-is-expectancy/

> Plain-text mirror for AI/LLM ingestion. Canonical HTML page: https://theoptionsbench.com/what-is-expectancy/

---

Updated 19 September 2026 · by Theo Chen

**Expectancy** is what one more trade is worth, on average, in dollars:
**(win rate x average win) - (loss rate x average loss)**. A win rate is
only one of those two terms, and on its own it cannot tell you whether a strategy makes money. In the
backtest below, the strike that won most often was also the one paid least to carry the worst losses.

Know your own win rate and average win and loss? The position sizing calculator turns them into a
stake per trade - it needs your win rate, your average win, your average loss and your account
size, which is the practical reason both halves of this page matter.

Open the Position Sizing Calculator ->

## A win rate is half a sentence

"I win 95.8% of my trades" tells you how _often_, never how _much_.
Both halves are needed, and in the test below they moved in opposite directions: the further out of the
money the strike, the more often it won and the less it was paid - and at the far end the average
loss was the largest of the seven, not the smallest. Two traders with identical win rates can run
opposite businesses.

This is not a hypothetical. Our
short strangle backtest sold the same structure at seven
different strikes over the same 14 years on SPY, 2010-2023 -
167 monthly cycles, except the 40-delta at 166.
One window, one structure, one variable. Read the win-rate column
and the profit column together:

| Strike | Win rate | Avg win | Avg loss | Loss / win | Expectancy | Total, 14y |
| 5 | 95.8% | $61 | $1,160 | 19.2x | +$9 | $1,567 |
| 10 | 85.6% | $120 | $625 | 5.2x | +$13 | $2,164 |
| 16 | 78.4% | $197 | $657 | 3.3x | +$13 | $2,155 |
| 20 | 73.7% | $253 | $642 | 2.5x | +$17 | $2,871 |
| 25 | 70.1% | $306 | $664 | 2.2x | +$16 | $2,601 |
| 30 | 67.1% | $350 | $670 | 1.9x | +$14 | $2,276 |
| 40 | 61.4% | $430 | $651 | 1.5x | +$13 | $2,164 |

Per contract. Average loss is not stored by the study; it is solved out of the identity
_total = wins x average win - losers x average loss_, so it carries the rounding
of those fields and is shown to the dollar. Expectancy is the formula applied to those parts, and it
reproduces the study's own published per-trade average to the cent - the build fails if it stops doing so.

## The highest win rate earned the least

The 5-delta won **95.8%** of all
167 months - the highest win rate of the seven - and finished with
**$1,567**, the _lowest_ total of the seven. The
20-delta won only 73.7% and finished with $2,871.
Ranking the strikes by win rate does not rank them by money: it puts the most reliable winner last.

The table says why in one column. At 5 delta the average loss is about
**19.2x the average win**. That is the trade being made: you are
buying a prettier win rate by accepting a worse payoff on the rare occasions you are wrong.
Only 7 of 167 months lost money and
they still cancelled about 84% of everything the
160 winning months made.
The study's own framing: a single worst month at that strike wiped out about
110 average winners.

## And a lower win rate earned more

The other direction is just as real, and it is the half people forget. The
30-delta won 67.1% of its months - it lost money in roughly
three months in ten - and earned **$13.63 a trade**. The
10-delta won 85.6% of the time, 18.5 percentage
points more often, and earned **$12.96**. The strike that won far less often
made _more_ per trade. Both ran the same 167 cycles, so that is a like-for-like
comparison rather than an artefact of one arm trading more often than the other. (The study records
each arm's cycle count, not its cycle dates, so that rests on the counts matching.)

It survives losing far more often because it is paid far more each time it is right: its average loss
is only 1.9x its average win, against 19.2x at the
far end. That is the whole mechanism. **Win rate and payoff are bought with the same currency,
and expectancy is what you have left after the exchange.**

## What a positive expectancy still does not promise

Every strike in that table has a positive expectancy, and it would be a mistake to read the winner as
the strike to sell. The entire spread between best and worst is $1,304, while the
standard error on any single strike runs from $6,729 to $12,608 -
the noise runs from about 5.2 to
9.7 times the signal. The study's own paired tests agree: the
largest statistic across the 4 strike pairs it tests is 0.55, below the
2 that would mark a difference worth believing - so on the pairs the study
checked, **no strike is measurably better than another**. An expectancy estimated from a finite
run of trades is itself an estimate, with error bars, and a positive one can be indistinguishable
from zero.

## How to use it on your own trades

- Take your last fifty closed trades from your broker's history, not your memory. Remembered trades flatter you, and a closed-only record flatters you too if losers are still sitting open - which is the mistake at the end of this page.
- Split them into winners and losers. Average each group separately, after commissions and after the spread you actually paid - not the mid you were quoted.
- Apply the formula. A negative answer means the strategy has been losing money regardless of how the equity curve felt.
- Then size it. A negative expectancy is a reason to stop; a positive one on fifty trades is a reason to keep measuring, not proof of an edge. Either way the same numbers - win rate, average win, average loss, and your account size - are what the position sizing calculator needs, and the Kelly simulator shows what a run of those trades looks like with the variance included.

One habit does most of the work here: keep the records. Expectancy computed from fifty real closed
trades is worth more than any rule of thumb, and it cannot be computed at all from a memory of how the
year went. That is the case for keeping a trading journal.

## Common mistakes

- Quoting a win rate as if it were a result. It is one of the two inputs, and it ranks against the money more often than with it: sorting the table above by win rate puts 14 of its 21 pairs in the wrong order by expectancy, the most reliable winner last.
- Letting losers stay open. Closed trades are the right input, but only if you close them. Running winners to a profit target while an unrealised loss sits open indefinitely keeps that loss out of the average, which is how a losing strategy shows a positive expectancy for years. Count what you are still holding before you trust the number.
- Ignoring the worst case because it is rare. The average loss in that table is an average. At 5 delta the worst single month was -$6,632, 5.7x the average loss, and it is the number that decides whether you survive to collect the expectancy at all.
- Treating a positive number as a proven edge. Fifty trades is enough to compute an expectancy. Whether it is enough to tell a real edge from luck depends on how big the edge is and how noisy the results are - which is the same question the spread-versus-error section above asks.

The bottom line

Expectancy is (win rate x average win) minus (loss rate x average loss), and a win rate is only one of those two terms - which is why a strike that won 95.8% of months in our own SPY test still earned the least of the seven we tested.

## Frequently asked questions
What is expectancy in trading?

Expectancy is the average profit or loss you can expect per trade, worked out from how often you win and how much you win and lose: (win rate x average win) minus (loss rate x average loss). Positive expectancy means the strategy makes money over enough trades; negative means it bleeds, however good the recent run looks. A win rate on its own cannot tell you which, because it is only one of the two terms.

Can a high win rate have negative expectancy?

Yes, and it is a common trap in options selling. Winning often says nothing about size. If you win $60 nineteen times out of twenty and lose $1,200 the twentieth, you win 95% of the time and still lose money. In our own SPY short-strangle test the 5-delta strike won 95.8% of months and earned the least of the seven strikes tested, because its average loss was about 19 times its average win.

Can a low win rate have positive expectancy?

Yes. A strategy that loses four times in ten can be perfectly sound if the six winners are bigger than the four losers. Trend following and long options both live here. In the same SPY test the 30-delta strike won 67.1% of its months against the 10-delta's 85.6%, over the same 167 cycles, and still earned more per trade - because it was paid more each time it was right.

How do you calculate expectancy?

Take your win rate as a decimal, multiply by your average winning trade, then subtract the loss rate multiplied by your average losing trade. A 60% win rate with $100 average wins and $120 average losses gives (0.6 x 100) - (0.4 x 120) = $12 per trade. Use real closed trades from your own records, not estimates, and include commissions and slippage.

Is expectancy the same as profit factor?

No, though they agree on the question that matters most. Profit factor is gross wins divided by gross losses, a ratio with no units; expectancy is dollars per trade. A profit factor above 1 and a positive expectancy always travel together, so either will tell you the historical edge was positive. Where they differ is what else they carry: expectancy is denominated in money, so it tells you what one more trade is worth, while the ratio does not. Neither is what a sizing rule takes - Kelly wants a win probability and a payoff ratio, and a calculator wants your account size on top of that.

## Related questions

- What did a high-win-rate short strangle actually earn?
- How big should each position be, given my edge?
- What does delta say about my chance of winning?
- How do I read the profit and loss curve of a trade?