Minimum Variance Delta

Black-Scholes delta answers the question "how many shares hedge me against the stock moving, holding implied volatility fixed?" But in equity markets, volatility does not sit still while the stock moves — when the stock drops, implied volatility reliably rises, and when it rallies, volatility tends to ease. An option's price responds to both moves at once, and a hedge built only against the stock, ignoring the vol move that historically comes with it, is quietly wrong on both the size and sometimes even the direction of the correction it should make.

Steering a car that also drifts sideways when you brake

Imagine a car where braking causes it to drift slightly sideways as an unavoidable side effect of the mechanics — a real, repeatable coupling, not random noise. A driver who only corrects the steering for forward speed, ignoring the sideways drift that braking always produces, will consistently under- or over-steer. A driver who has learned the coupling corrects for both at once, using one combined steering adjustment. Minimum variance delta is that combined correction: instead of hedging only against the stock, it hedges against the stock move and the volatility move that empirically tends to accompany it.

The adjustment

Minimum variance delta starts from the ordinary Black-Scholes delta and adds a correction for the empirical relationship between the stock's move and implied volatility's move:

$$\Delta_{\text{MV}} \;=\; \Delta_{\text{BS}} \;+\; \text{vega} \times \frac{\partial \sigma_{\text{imp}}}{\partial S}.$$

In plain English: take the ordinary delta, then add the option's vega multiplied by how much implied volatility typically moves for a given move in the stock — the "skew slope." Because equity skew is downward-sloping (a stock drop is empirically paired with a volatility rise, so $\partial \sigma_{\text{imp}}/\partial S$ is negative), and vega is always positive, this correction is negative for both calls and puts: minimum variance delta sits below Black-Scholes delta.

Worked example 1: quantifying the correction

A one-year at-the-money put, stock at $100, $\sigma=20\%$, Black-Scholes delta $\Delta_{\text{BS}} = -0.46$, vega $= 0.39$ per volatility point. Suppose the observed skew slope is $-0.5$ volatility points per $1 move in the stock (a common order of magnitude for equity index skew): $\partial\sigma_{\text{imp}}/\partial S = -0.005$ per $1, i.e., $-0.5$ (vol points) per $1.

$$\Delta_{\text{MV}} = -0.46 + (0.39)(-0.5) = -0.46 - 0.195 = -0.655.$$

The minimum variance delta calls for hedging as though this put were significantly more sensitive to the stock than Black-Scholes says — because empirically, whenever this stock falls, it drags implied volatility up with it, and the higher volatility adds extra value to the put on top of the pure stock-price effect.

Worked example 2: the hedge error each formula leaves behind

The stock drops $2, and true to the historical pattern, implied volatility rises by 1 point (from 20% to 21%). The put's actual price change has two pieces: the delta effect plus the vega effect from the vol rise.

Actual move (approximately): $\Delta_{\text{BS}} \times (-2) + \text{vega} \times (1) = (-0.46)(-2) + 0.39(1) = 0.92 + 0.39 = 1.31$.

Predicted by Black-Scholes delta alone (no vol move accounted for): $(-0.46)(-2) = 0.92$. Error: $1.31 - 0.92 = 0.39$, an undershoot of 39 cents purely because Black-Scholes delta has no way to see the vol move coming.

Predicted by minimum variance delta: $(-0.655)(-2) = 1.31$. Error: $1.31 - 1.31 = 0.00$. By construction, the minimum variance delta already has the empirical vol-move baked in, so it tracks the put's actual move almost exactly for the typical, historically-observed size of vol move that accompanies a $2 drop.

The downward tilt visible below, low strikes trading at higher implied volatility than high strikes, is exactly the empirical relationship minimum variance delta bakes into its correction. Explore how the surface's slope varies by tenor — the steeper the local slope, the bigger the gap between Black-Scholes delta and minimum variance delta.

What this means in practice

Equity index option desks routinely hedge with a minimum variance delta rather than a raw Black-Scholes or even local-volatility delta, because the correction measurably reduces day-to-day hedging P&L noise for skew-heavy underlyings like equity indices, where the stock-vol relationship is strong and stable. It matters most for out-of-the-money puts, where vega is still meaningful and the skew slope is steepest.