Please update Google Chrome

Premium Tracker needs a newer version of Chrome to display correctly. Update Chrome from the Play Store, then reopen the app.

Update Chrome
University
402 · Performance & Metrics

A 90% Win Rate Tells You Almost Nothing

Selling further out raises the win rate and lowers the expected value simultaneously.

Advanced12 min readUpdated

Worth reading first: Probability of Profit, and Why It Is Not Enough

Course catalogue99 lessons

Premium selling hands you a 90% win rate on day one. You do not have to be good at it, pick the right stocks, or time anything. Sell far enough out of the money and the win rate arrives automatically.

Which is exactly why it is worthless as a measure of whether you are any good at this.

Win rate and expected value, moving in opposite directions

Drag the strike down. Watch the probability of profit climb toward 95% while the expected value shrinks. The two numbers are not measuring the same thing.

$93.00
30 days
30%
  • Where the stock is likely to land
  • What you make or lose there
Probability of profit83%
Expected value−$0every outcome, probability-weighted
Max profit$86the credit
Worst case−$9,214stock to zero

Push the strike far enough down and the probability of profit climbs toward 95% while the expected value shrinks. A high win rate and a poor trade are entirely compatible, the two numbers are answering different questions.

A lognormal model, which understates real tail risk. The divergence shown here is the optimistic version.

Expectancy is the number that decides

Expectancy = (win rate × average win) − (loss rate × average loss)

Win rate is one of four inputs, not the answer.

A strategy is profitable when expectancy is positive. Win rate on its own tells you nothing about that, because it says nothing about the sizes.

Win rateAvg winAvg lossExpectancy per trade
95%$40$1,200−$22
90%$80$700+$2
75%$180$400+$35
55%$400$300+$85

Four strategies. The win rate column ranks them exactly backwards.

The 95% strategy loses money. The 55% strategy is the best of the four. Anyone selecting by win rate would pick precisely the wrong one, and would feel excellent about it for a long time before finding out.

Why premium selling maximises the divergence

Every short option has a capped gain and an uncapped or much larger loss. That is the definition of the position, not a risk you took on badly.

So the average win is structurally small and the average loss is structurally large, which means the win rate has to be high just to break even. Moving the strike further out raises the win rate and shrinks the credit at the same time. The two effects push expectancy in opposite directions and there is no strike where both improve.

Win rate ↑ ⟺ average win ↓, at every strike

The relationship you are trading along, not escaping.

Measuring your own expectancy

It needs four numbers from your records: the count of wins and losses, and the average size of each. All four come from a properly kept log, and three of them are unavailable if you record only premium collected.

Two cautions when you do it. First, include unrealised losses on assigned shares, excluding them turns losing trades into open positions and inflates the win rate directly. Second, a year is not enough data. Premium selling's losses cluster, and a sample without a bad quarter in it will overstate your expectancy substantially.

A better pair of questions

Rather than “what is my win rate?”, two questions that actually inform:

How many wins does one loss cost me? If the answer is fifteen, your win rate needs to exceed 94% before costs, which is a far more demanding target than it sounds.

What does my worst realistic month do? Expectancy assumes you survive to collect the average. See drawdown and risk of ruin.

What can go wrong

Selecting strikes by win rate. Drives you toward negligible expected value.

Excluding assigned shares from the loss count. Inflates the win rate mechanically.

Sizing up after a streak. The streak was structural.

Judging expectancy on one good year. Losses cluster; a quiet sample flatters.

Key takeaways

  1. A 90% win rate arrives automatically with out-of-the-money selling and measures nothing about skill.
  2. Expectancy is win rate, loss rate and both average sizes together. Win rate alone can rank strategies backwards.
  3. Premium selling has capped wins and large losses by construction, so a high win rate is required merely to break even.
  4. Raising the win rate shrinks the credit; there is no strike where both improve.
  5. Include unrealised losses on assigned shares, and do not judge expectancy on a sample with no bad quarter in it.

Check your understanding

  1. 1. 95% win rate, $40 average win, $1,200 average loss. What is the expectancy?

  2. 2. Why does raising your win rate usually lower your expectancy in premium selling?

  3. 3. Which record-keeping error mechanically inflates a reported win rate?