Picking Strikes by Delta for a "Target Win Rate": Why the Real Number Runs Lower
The heuristic everyone learns first
Ask any options trader how they pick a strike for a credit spread and you'll hear a version of the same rule: sell the 30-delta strike for roughly a 70% win rate, sell the 16-delta strike for roughly 84%, sell the 10-delta for roughly 90%. It's a genuinely useful shortcut — delta is sitting right there on the chain, no separate calculation required, and it scales cleanly across any underlying or expiration. The problem isn't that the shortcut is useless. It's that traders quietly upgrade "delta is correlated with probability" into "delta is the probability," and those are not the same claim.
Where the number actually comes from
As covered in more depth in delta as a probability proxy, a call's delta under Black-Scholes is N(d1), while the model's own risk-neutral probability of finishing in the money is a slightly different term, N(d2). They converge as volatility gets small and drift apart as it rises. That's already a crack in the "delta equals probability" claim before real markets even enter the picture: the two numbers were never meant to be identical, just close cousins that share a formula.
More importantly, both N(d1) and N(d2) are risk-neutral probabilities — they come from a pricing model that assumes the stock drifts at the risk-free rate and moves in a lognormal way around it. Nothing in the option chain is measuring how the stock will actually behave. It's measuring a mathematical probability consistent with today's price, under assumptions that are convenient for pricing, not guaranteed to match reality.
The two gaps between model and outcome
Two separate effects push realized win rates below the delta-implied number, and they compound rather than cancel out.
The first is the volatility risk premium: implied volatility tends to run above what the stock subsequently realizes, because buyers of options are willing to overpay for protection against tail moves. That sounds like good news for a seller — and on average it is, since it's the reason selling premium has a structural edge at all. But it doesn't change the shape of the distribution the strike sits against; it just means the market's estimate of the whole distribution's width was pessimistic more often than not. It's a return-generating effect, not a win-rate-boosting one, and treating it as the latter overstates how safe a given strike really is.
The second, more direct effect is volatility skew. Delta is computed using the implied volatility quoted for that specific strike, not a single flat number across the chain. Because downside strikes trade at richer implied vol than the lognormal model assumes is "fair," the market is already pricing fatter left-tail risk than a symmetric bell curve would predict. The delta figure you read off the chain is closer to reality than a textbook Black-Scholes number with one flat IV input — but the underlying lognormal assumption still understates how often genuinely large, fast down-moves occur relative to a distribution with true fat tails.
Frequent small wins, infrequent large losses
This is why premium-selling strategies built around a fixed target delta tend to produce a specific return pattern: a long stretch of trades that close near max profit, punctuated by occasional losses that are disproportionately large relative to the credit collected. Each individual loss doesn't have to be common to matter — it just has to arrive more often, or land harder, than the "walk-away probability" implied by the entry delta suggested it would. As explored in iron condor risk-reward math, the payoff asymmetry of a short-premium position means the tail event doesn't need to be frequent — it just needs to be under-priced relative to the strike selection, and fat tails plus skew are exactly the mechanism that keeps quietly under-pricing it.
Using delta as a screen, not a promise
None of this means delta-based strike selection is a bad starting filter — it's a fast, consistent way to compare relative distance-from-the-money across different underlyings and expirations. The fix is in how much weight the number is given afterward. Treat the delta-implied win rate as an upper bound rather than an expectation, check where current implied volatility sits using IV rank and IV percentile before assuming today's premium is fairly compensating for the risk, and look at the skew shape directly rather than only the single strike you're considering — a steep skew is the chain telling you, in its own units, that the tail you're selling against is priced fatter than a flat-vol delta number would suggest.
You can see delta, IV, and skew together on a live chain in the Greeks Matrix, or brush up on the underlying definitions in the glossary.