Delta, Volatility, and the Two Faces of d₁

Why a book's deltas cluster near 0.5 as you raise volatility — even though a call's delta tends to 1 as $\sigma \to \infty$.

The apparent contradiction

Two statements, both correct, that seem to fight each other:

• In Black–Scholes, as volatility becomes infinite, a call's delta tends to 1.
• On a real trading book, when you raise the volatility input, your deltas do the opposite — they compress toward 0.5.

There is no contradiction. The two describe different regimes, and both live inside a single expression.

One equation settles it

With $r=q=0$ and a flat surface, the call delta is $\Delta = N(d_1)$, where

$$d_1 \;=\; \underbrace{\frac{\ln(S/K)}{\sigma\sqrt{\tau}}}_{\text{moneyness term}} \;+\; \underbrace{\frac{\sigma\sqrt{\tau}}{2}}_{\text{variance-drift term (Itô } \tfrac12\sigma^2)} .$$

Everything follows from how these two terms behave as $\sigma$ grows.

Regime 1 — your book (realistic vols): the moneyness term wins

Raising $\sigma$ drives the first term $\ln(S/K)/(\sigma\sqrt{\tau})$ toward $0$: the strike's distance from spot becomes negligible next to the width of the distribution. Every option "looks at-the-money," so $d_1 \to 0$ and its delta is pulled toward $N(0)=0.5$. Out-of-the-money deltas rise toward it; in-the-money deltas fall toward it. That compression is what you see on the book.

The second term is tiny at realistic vols, so it barely lifts the cluster above 0.5. For an at-the-money option, $\Delta = N(\sigma\sqrt{\tau}/2)$:

Your book lives at the top of that table: deltas around 0.54–0.60, "stuck" near 0.5.

Regime 2 — infinite volatility: the drift term wins

The variance-drift term $\sigma\sqrt{\tau}/2$ grows without bound and eventually pushes $d_1\to+\infty$, so $\Delta\to 1$ and the call price $\to S$. But it grows only like $\sigma\sqrt{\tau}$, so it overtakes the moneyness term only at absurd volatilities — far beyond anything a book ever carries. "Delta $\to 1$" is a limit, not a working regime.

The subtlety worth knowing

As $\sigma\to\infty$, something beautiful happens:

$$N(d_1) \to 1 \quad (\text{delta}), \qquad N(d_2) \to 0 \quad (\text{risk-neutral probability of exercise}),$$

because $d_2 = d_1 - \sigma\sqrt{\tau} \to -\infty$. The call almost surely finishes worthless, yet its delta tends to 1 and its value tends to $S$. The reconciliation is the lognormal's split personality: its median $S\,e^{-\sigma^2\tau/2}\to 0$ while its mean stays at $\mathbb{E}[S_\tau]=S$. All the value migrates into an ever-rarer, ever-larger right tail. The call becomes the stock itself — downside capped at $0$, strike irrelevant — so delta $\to 1$.

Which deltas move which way

Same story per strike (calls, $r=q=0$):

• OTM ($S<K$): delta rises monotonically, $0 \to 0.5 \to 1$.
• ATM ($S=K$): $\Delta=N(\sigma\sqrt{\tau}/2)$, always slightly above 0.5, creeping up.
• ITM ($S>K$): delta dips from 1 toward — but never below — $\sim$0.5, then climbs back to 1.

Puts are the mirror image: $\Delta_{\text{put}} = N(d_1)-1$, and the absolute deltas compress the same way.

An everyday picture

Think of a bet where the most you can lose is a small fixed stake, but your winnings are unlimited — that is a call. How much does your wealth ride on the outcome? When the possible swings are mild, only about half of it does: a good result gains you a bit, a bad one just costs the stake — it feels like a coin flip on whether you end up ahead (your 0.5). But let the swings get wild — the outcome could be 100× or nothing — and the fixed stake becomes a rounding error next to the upside: essentially every euro of your result now rides on the bet, so you effectively own the whole thing (delta 1). That extreme regime is real, but you only ever see the half-exposed number — exactly like your book.

Takeaway. The "$\Delta\to 1$" comes from the $\tfrac12\sigma^2\tau$ drift term, which dominates only at infinite volatility. On any real book it is the moneyness term that governs: raising vol washes out the distance to the strike and pulls every delta toward 0.5.

Assumes $r=q=0$ and a flat volatility surface (the exam convention). Non-zero carry shifts the two terms but not the mechanism.