Probability of Exercise and Dual Delta

Traders routinely say "that option has a 30% chance of finishing in the money" while looking at a Greek called delta, officially defined as the option price's sensitivity to a $1 move in the stock — not a probability at all. The shorthand isn't wrong so much as imprecise: buried inside the Black-Scholes formula is a genuine probability, but it isn't quite delta. It's a close cousin called dual delta, and the difference matters whenever the two numbers pull apart, which is most of the time.

Think of a dartboard where the bullseye is "stock ends above the strike." Dual delta is the fraction of the future stock price's probability distribution that lands past the strike — shade the area of the bell curve to the right of the strike line, and that shaded fraction is dual delta. Delta is different: how fast the value of the dart game changes as the whole board shifts sideways. A shift changes both where darts land and how much a winning dart is worth, and delta bundles both effects into one number, while dual delta isolates just the first.

The two formulas side by side

For a European call under Black-Scholes:

$$\Delta = N(d_1), \qquad \Delta_{\text{dual}} = N(d_2) = e^{r T}\frac{\partial C}{\partial K}\Big/(-1) .$$

$N(\cdot)$ is the cumulative normal distribution — the area under a bell curve up to some point. $d_1$ and $d_2$ are the two standard Black-Scholes terms, with $d_2 = d_1 - \sigma\sqrt{T}$, so $d_2$ is always a little smaller than $d_1$. $N(d_2)$ is, in plain English, the risk-neutral probability the stock finishes above the strike at expiry. $N(d_1)$ is delta — not that probability, even though the formulas look nearly identical; it's the probability-weighted answer to "how much does the option's payoff shift," mixing in that a higher stock price also means a bigger payout. The gap between $N(d_1)$ and $N(d_2)$ is exactly the volatility-time adjustment $\sigma\sqrt{T}$ tucked inside $d_1$.

Worked example 1 — reading off the probability

A stock trades at $100, a 3-month call is struck at $105, volatility is 25% annualized, and the risk-free rate is 3%. The standard Black-Scholes terms give $d_1 \approx -0.09$ and $d_2 \approx -0.22$. Looking these up: $N(d_1) \approx 0.464$ — delta, so the option's price moves about $0.46 for every $1 move in the stock. $N(d_2) \approx 0.412$ — dual delta, the actual risk-neutral probability the stock finishes above $105 in three months, about 41%. A trader who quotes "that's a 46-delta call" and treats it as "46% chance of exercise" is off by about five points, because delta overstates the exercise probability whenever volatility is meaningfully above zero.

Worked example 2 — the gap widens with volatility

Hold everything else fixed and raise volatility from 25% to 60%. Now $d_1 \approx 0.03$ and $d_2 \approx -0.27$, giving $N(d_1) \approx 0.512$ and $N(d_2) \approx 0.394$. Delta barely moved (46.4% to 51.2%) but dual delta fell further, to 39.4%. The gap — 5 points at 25% vol, nearly 12 points at 60% vol — grows directly with $\sigma\sqrt{T}$. The lesson: the higher the volatility, the less delta should be trusted as a stand-in for "probability of finishing in the money," because delta is inflated by the extra upside a volatile stock can deliver given it finishes there, while dual delta only cares whether it crosses the line at all.

Picture the future stock price (in log terms) as the bell curve here, with the strike marked as a vertical line along the x-axis. Dual delta is exactly the shaded area to the right of that line. Drag the standard deviation up and watch the curve flatten — more of the distribution spreads past the strike but also further beyond it, which is why raising volatility doesn't move dual delta and delta by the same amount.

What this means in practice

Traders use dual delta directly to build digital options (pay $1 if the stock finishes above the strike, nothing otherwise) — its price is simply $e^{-rT} N(d_2)$, the discounted probability, with no payoff-size adjustment needed since the payout is a flat dollar amount. It also shows up whenever someone needs a genuine probability estimate implied by the option market — for instance, estimating the market-implied odds of a takeover price being exceeded — rather than a hedge ratio.

Practice

1. For a put option, dual delta is $-N(-d_2)$ (in absolute value, $N(-d_2)$). Using the 25%-volatility numbers from worked example 1, roughly what's the risk-neutral probability that the $105 strike finishes out of the money for a call — and how does that relate to the put's dual delta at the same strike?
2. Why does raising time-to-expiry, holding volatility fixed, widen the gap between delta and dual delta in the same way that raising volatility does?