What options implied probability is
Options implied probability is the probability the option market charges for a payout. If a position that pays one dollar when SPX settles above 7,700 on Friday costs 23 cents, then the chain is pricing those odds at 23 percent. That is the whole idea. It is a price quoted as a percentage, not a survey of opinion and not a count of past outcomes.
The number is worth having because it is the only forward-looking probability in markets that is both public and continuously updated. Traders, banks and funds transact against it with real money all day, and the chain publishes the result. You can read it off the screen in about ten seconds once you know where to look, and our implied odds page reads it off the delayed Cboe chain for SPX, SPY and QQQ.
How to read it off a chain in ten seconds
The arithmetic is a subtraction and a division.
Find the two listed strikes either side of the level you care about. Take the call price at the lower strike, subtract the call price at the higher strike, then divide by the distance between the strikes. The answer is the probability the market prices on settling above the level between them.
Work a real example. Suppose SPX calls expiring Friday are quoted at 15.20 for the 7,690 strike and 13.60 for the 7,700 strike. The difference is 1.60 across ten points of strike, so 1.60 divided by 10 is 0.16. The chain is pricing a 16 percent probability that SPX settles above roughly 7,695 on Friday.
That construction is called a digital spread, or a call spread scaled to pay one dollar. Buying the 7,690 call and selling the 7,700 call costs 1.60 and pays a maximum of 10, so you are paying 16 cents per dollar of payout. The fraction is the probability, and the reason it is the probability is that a bet paying one dollar when an event happens is worth exactly the odds the market puts on it.
Where the formula comes from
Douglas Breeden and Robert Litzenberger proved the general version of this in 1978. Their result is that the entire probability distribution the market is pricing can be recovered from the curve of call prices against strike. The first derivative of that curve gives you the probability of settling above any level. The second derivative gives you the probability density itself, which is the full shape of what the market expects.
The digital spread above is the first derivative, approximated across two adjacent strikes. Everything else in this space is a refinement of that one step: smoothing the curve before differentiating it, handling strikes that are too far apart, and deciding what to do when the quotes misbehave.
Risk-neutral is not real-world, and the gap is large
This is the part that separates a useful reading from a misleading one.
The probability you extract from option prices is called risk-neutral. It is the probability that makes today's prices fair if nobody demanded any compensation for bearing risk. But investors do demand compensation, particularly for downside. Portfolio managers pay above fair odds for crash protection the way homeowners pay above fair odds for fire insurance, because the loss they are insuring against is the one that hurts most.
The consequence is systematic and one-directional. The priced probability of a large decline sits above the frequency with which declines of that size have actually happened. The priced probability of a large rally sits below. Academic work on this gap, starting with the index-option pricing literature of the late 1990s, consistently finds the risk-neutral distribution has a fatter left tail than the realised one.
So when a chain prices a 6 percent probability of a five percent fall by Friday, the correct reading is "protection against a five percent fall costs six cents on the dollar." The incorrect reading is "there is a six percent chance of a five percent fall." The first statement is a measurement. The second is a forecast the chain never made.
Treat the number as the cost of a view, not the likelihood of an outcome. That framing is what makes it actionable: it tells you whether a payout is cheap or expensive relative to what you think, which is a comparison you can act on, rather than handing you a prediction you cannot check.
Where this sits next to the tools you already use
Three neighbouring numbers get confused with implied probability, and each answers something different.
The expected move is a symmetric band around the current price built from one at-the-money volatility. It answers "how far" in both directions at once. Implied probability answers "how likely, above this specific level", which is a different question and, crucially, an asymmetric one. Skew in the chain means the priced odds of a five percent fall and a five percent rise are rarely mirror images, and a symmetric band cannot show you that.
Delta is the other shortcut, and it is the closest of the three. A 0.30 delta call sits near a 30 percent probability of finishing in the money, which is why the approximation is taught. But delta is a hedge ratio, not a probability: in the Black-Scholes framework delta is N(d1) while the probability of finishing in the money is N(d2), and the two separate as volatility and time to expiration grow. The greeks explainer works through what delta actually measures.
Implied volatility is the input, not the output. A single implied volatility reading describes the whole distribution as if it were symmetric and lognormal, which it is not. The digital spread reads the probability directly from prices instead, so the skew is already inside the answer rather than being assumed away.
When the reading is not trustworthy
Four conditions break the arithmetic, and all four are visible on the chain before you compute anything.
Strikes too far apart. The difference across two strikes is an average of the probability over that whole interval. On an index where the listed strikes are twenty-five points apart and you want the odds above a level between them, the answer is a local average rather than a reading about your level. The wider the gap relative to the underlying, the less the number means.
Wide or stale quotes. A midpoint between a bid and an ask nobody transacted at is not a price. When the spread across two adjacent strikes is comparable to the price difference between them, the subtraction is measuring the spread rather than the market.
In-the-money quotes. Deep in-the-money contracts are quoted badly on delayed feeds because nobody is trading them. Any careful construction uses out-of-the-money options on both sides, converting puts to calls with put-call parity below the forward, which is what we do on the implied odds page.
Long horizons. Past a few months the discount factor, the dividend assumption and the bid-ask width all stop being rounding errors. A probability quoted out to next year carries assumptions the chain did not supply.
If any of those hold, the honest answer is no number rather than a smoothed one.
How to use it
Two uses justify the effort.
The first is comparing your own view to what you would be paying for it. If you think a level is a coin flip by Friday and the chain prices it at 20 percent, that is a concrete disagreement with a price attached, and you can decide whether it is one you want to fund. If you think it is a long shot and the chain prices it at 45 percent, you have learned something about positioning before you commit anything.
The second is reading how the priced distribution shifts. A jump in the priced odds of a downside level, with the corresponding upside level unchanged, tells you protection got bid without spot moving. That is a positioning observation about the option market, which is the same class of information as the dealer gamma picture, and it is dated and checkable in a way that commentary is not.
What it will not do is tell you what happens next. The chain has no more idea than you do. What it has is a price, and a price you can read is worth more than an opinion you cannot.
Educational content, not financial advice. See our risk disclosure.