Delta Hedging Frequency And Costs

A driver steering a car down a straight road doesn't hold the wheel rigidly still, and doesn't yank it every tenth of a second either. Small, occasional corrections keep the car centered; correcting too rarely and the car drifts to the shoulder before you notice, correcting too often just wears out the steering linkage for no benefit. Delta hedging an option book is the same problem with dollars instead of a steering wheel. The textbook version of Black-Scholes assumes you rehedge continuously, every instant, for free. No real desk does that — every trade to adjust the hedge costs a bid-ask spread and a commission — so the actual decision a trader makes every day is: how often is often enough?

Why the hedge needs adjusting at all

Delta is the option's sensitivity to the stock price, and it isn't constant — it changes as the stock moves, because the option's value curve is bent (that bend is gamma). A hedge that was exactly right this morning is slightly wrong by lunchtime, because the stock moved and delta moved with it. Rehedging means buying or selling shares to bring your stock position back in line with the option's new delta. Do it continuously and — in a frictionless world — the randomness cancels out perfectly, which is the whole point of the Black-Scholes derivation. Do it in discrete jumps, which is all anyone can actually do, and a little bit of noise survives each interval.

The noise per interval, in one formula

Between rehedges, the option and the (now slightly stale) hedge drift apart by an amount driven by gamma:

$$\text{P\&L over one interval} \;\approx\; \tfrac{1}{2}\,\Gamma\,(\Delta S)^{2} \;-\; \theta\,\Delta t .$$

In plain English: $\Gamma$ ("gamma") measures how fast delta itself moves as the stock moves. $\Delta S$ is how far the stock actually moved during the interval, and $\Delta S$ squared shows up because the gain from gamma doesn't care which direction the stock moved — up or down, a bent curve gains either way. $\theta$ ("theta") is the option's time decay, the rent you pay every day just for holding it, and $\Delta t$ is the length of the interval. On average, if the stock's realized volatility matches what the option was priced with, these two terms cancel exactly — that's the whole promise of the Black-Scholes hedge. What doesn't cancel is the randomness of $(\Delta S)^2$ around its average: some intervals the stock barely moves and you collect theta for nothing in return; other intervals it jumps and gamma hands you a windfall. That leftover noise is the hedging error, and it is exactly what more frequent rehedging shrinks.

Worked example 1 — a single rehedge interval

A desk holds an option with gamma $\Gamma = 0.05$ per dollar of stock move (in per-share terms, scaled to a 100-share contract this is 5 shares of delta drift per $1 move), and theta of $8 per day on that contract. Over one day the stock moves $2. Gamma P&L: $\tfrac{1}{2}(0.05)(2)^2 = \tfrac12 (0.05)(4) = 0.10$ per share, or $10 on a 100-share contract. Subtract the day's theta cost of $8: net P&L for the interval is $10 − $8 = $2. If the stock had instead moved only $1: gamma P&L is $\tfrac12(0.05)(1)^2 = 0.025$ per share, $2.50 on the contract, against the same $8 theta — a net loss of $5.50 for that day. Same option, same hedge discipline, opposite sign of daily P&L purely because of how far the stock happened to move that day. Averaged over many such days it washes out; on any single day it doesn't.

Worked example 2 — frequency versus cost

The classic result (due to Boyle and Emanuel) is that the variance of your total hedging error over an option's life shrinks in proportion to $1/N$, where $N$ is the number of times you rehedge. Since standard deviation is the square root of variance, quadrupling $N$ halves the standard deviation of your hedging error.

Suppose rehedging weekly ($N = 52$ over a one-year option) leaves a hedging-error standard deviation of $2,000, and each rehedge trade costs $50 in spread and fees. Total transaction cost: 52 × $50 = $2,600. Now rehedge four times as often, $N = 208$: the standard deviation falls by half, to $1,000, but transaction cost rises to 208 × $50 = $10,400. Adding cost and risk together as a rough measure of total pain: weekly totals $2,600 + $2,000 = $4,600; four-times-weekly totals $10,400 + $1,000 = $11,400. In this example, hedging more often actually makes things worse overall — the linear growth in transaction costs outruns the square-root shrinkage in risk. There is a frequency in between that minimizes total pain, and finding it, not maximizing hedge accuracy, is the real optimization.

Watch a simulated stock path above. Each little wiggle is an interval where a discretely-hedged book picks up a small gamma gain or loss; a continuously-hedged book would erase every one of them instantly, at infinite cost.

What this means in practice

No real book rehedges on a fixed clock. Desks rehedge when delta drifts past a threshold (a "band"), rehedge more aggressively on high-gamma, high-notional books, and rehedge less on positions near expiry where gamma spikes are large but short-lived and thrashing the hedge just bleeds cost. The frequency decision interacts directly with position size: a $100,000 option book can afford to rehedge daily because the fixed cost per trade is small next to the gamma risk; a $100 million book might rehedge the same option type on a much tighter band because a single day's hedging error is now a serious dollar number.

Practice

1. If weekly rehedging ($N=52$) gives a hedging-error standard deviation of $3,000, what standard deviation would you expect from rehedging nine times as often ($N=468$), using the $1/\sqrt{N}$ scaling?
2. A desk pays $20 per rehedge trade instead of $50. Redo worked example 2's comparison (weekly vs. four-times-weekly) with this lower cost — does the conclusion about which frequency is better change, and why?