Delta-Neutral Hedging

You already know the five gauges. Delta tells you your stock exposure, gamma tells you how fast that exposure mutates, theta is the rent, vega the volatility bet. This lesson is where those gauges stop being trivia and start being a job: keeping a book delta-neutral isn’t a single trade you make and forget — it’s a chore you redo, over and over, fighting gamma the whole way, while transaction costs quietly bill you for every fidget. That tension — hedge often and pay the broker, or hedge rarely and pay the market — is the whole game.

Quick instinct check before we wire it up.

What it means to be delta-neutral

Analogy. Being delta-neutral is balancing a broom on your palm. You don’t set your hand once and admire it — the broom is always tipping, and you survive only by a constant stream of tiny corrective shuffles. Stop correcting and it falls. A delta hedge is the same: not a state you reach, but a state you continuously defend.

Definition & mechanics. A position’s delta is the sum of the deltas of everything in it, measured in equivalent shares of the underlying. You are delta-neutral when that total is zero: $\Delta_{\text{position}} = 0$. To first order — for a small move $dS$ — your profit and loss is $\text{P\&L} \approx \Delta_{\text{position}} \cdot dS$, so a zero delta means small wobbles in the stock don’t move your money. You’ve cancelled the directional bet and kept only the higher-order stuff (gamma, theta, vega).

The lever you pull to get there is the hedge ratio: delta is literally shares of stock per option. Hold an option with delta $\Delta_o$ on a 100-share contract and you carry $100\,\Delta_o$ shares of equivalent exposure; trade $-100\,\Delta_o$ shares of stock against it and the first-order exposure nets to zero.

Worked example. You short 1 call with delta $0.50$ on 100 shares. The short call’s position delta is $-0.50 \times 100 = -50$ — you behave like someone short 50 shares. To neutralise, buy 50 shares. Now stress it with the stock at $100:

• Stock rises to $101. Short call loses about $50 (since $0.50 \times 100 = 50$; the option you sold got more valuable). Your 50 long shares gain $50 (that’s $50 \times 1$). Net $\approx 0$.
• Stock falls to $99. Short call gains about $50; long shares lose $50. Net $\approx 0$.

The book barely flinches in either direction. That is exactly what neutrality buys you — for now.

Why the hedge drifts (gamma)

Analogy. You’ve zeroed the bathroom scale with a 5 kg weight on it. Add a second weight and the zero you so carefully set is meaningless — the calibration only held at that one load. Delta is a calibration that only holds at one stock price. Move the price and gamma silently re-calibrates your delta out from under you.

Mechanics. Delta is the slope of the option’s value curve, and that curve is curved — gamma is the curvature, $\Gamma = \partial \Delta / \partial S$. So as $S$ moves by $\Delta S$, your option’s delta shifts by roughly $\Gamma \cdot \Delta S$. The hedge you placed against the old delta is now wrong by that much. You must re-hedge: trade more stock to drag the position back to zero. Then the stock moves again, and you do it again. Forever, until expiry.

Worked re-hedge. Start from the book above: short 1 call, delta $0.50$, hedged with $+50$ shares, stock at $100. The stock rallies $5 to $105.

The simulator below makes the drift visible. The orange line is the naked short call — pure chaos, lurching with every tick. The blue line is the same call delta-hedged: drag the rebalance slider and watch how a hedge that updates more often flattens to near-zero, while a lazy hedge lets gamma slippage leak through.

Continuous vs discrete hedging

Analogy. Black–Scholes is a physics problem solved in a frictionless vacuum: it assumes you can rehedge continuously and for free. That’s like a textbook ramp with no friction — beautiful math, but you can’t actually build it. The real world has friction (costs) and you can only touch the position at discrete moments (you sleep, markets close, you can’t trade infinitely fast).

The continuous ideal. If you could rehedge continuously and costlessly, the hedge would track the option’s delta exactly at every instant, your replicating portfolio would perfectly shadow the option, and a fairly-priced option would net you exactly zero P&L. That’s the replication argument at the heart of Black–Scholes: the hedge is the option.

The discrete reality. You actually rehedge at $N$ discrete times. Between two rehedges the stock moves by some $\Delta S$ while your share count is frozen, so your linear (delta-only) hedge misses the curvature of the option’s payoff. Over one interval the leftover P&L — the hedging error — is approximately

The $\tfrac12\Gamma(\Delta S)^2$ term is the gamma P&L the straight-line hedge couldn’t capture; theta is the offsetting time-decay term. On a fairly-priced option these two roughly cancel on average, but on any single interval they don’t — and the leftover is your tracking error. Critically, the error per interval scales with gamma: the more curved the option, the more a frozen hedge mis-tracks it.

Add up the random per-interval errors over the life of the trade and the standard deviation of the total hedging error scales like $1/\sqrt{N}$. Rehedge four times as often and you roughly halve your hedging-error risk. Push $N \to \infty$ and the error vanishes — recovering the continuous, perfect-replication ideal. So far, more hedging is pure upside. (Then the bill arrives.)

Transaction costs and the U-shaped tradeoff

Analogy. Hedging error is like noise on a phone line: keep re-syncing and it gets quieter. But suppose every re-sync charged you a toll. Now syncing constantly gives you a crystal-clear line and a monstrous phone bill. Somewhere between “never sync” (unusable) and “sync constantly” (broke) is the sweet spot. Hedging lives in exactly that valley.

Mechanics. Every rehedge is a real trade: you cross the bid-ask spread and may pay a commission. So while rehedging more often shrinks the hedging-error std-dev (the $1/\sqrt{N}$ from before), it grows your total transaction cost. Because each rehedge over a fixed horizon moves delta by only about $1/\sqrt{N}$, the cumulative turnover cost scales like $\propto \sqrt{N}$ (a simpler model just charges a flat fee per trade and gets $\propto N$). Either way, cost rises with $N$ while error falls with $N$:

Add a falling curve to a rising curve and you get a U-shape: the total cost drops, bottoms out, then climbs. The bottom of that U is the optimal rehedging frequency — hedge any more or any less and your total cost goes up. This is the single most important practical fact about dynamic hedging.

Drag the transaction-cost slider below. Watch the optimum (the dot at the bottom of the blue total-cost curve) slide toward fewer rebalances as trading gets pricier.

Worked tradeoff example

Take a quarter-long short-call hedge and compare three rehedging regimes. Suppose the hedging-error std-dev is $1,000 at $N=1$ (so it scales as $1000/\sqrt{N}$) and turnover cost runs about $60 at $N=1$ (scaling as $60\sqrt{N}$).

Reading down the table: as you rehedge more, the error column keeps falling (good) but the cost column keeps rising faster (bad). Hourly hedging nearly zeroes the tracking error yet triples the total bill versus weekly. The minimum here sits around weekly rehedging — the U-curve’s valley — not at the extreme. “Continuous-ish” hedging, the Black–Scholes ideal, is a financial disaster in a world with spreads: perfect tracking at infinite cost. The art is finding the bottom of the U, not racing toward zero error.

Gamma and vega still leak

Hedging delta to zero and rehedging diligently buys you freedom from small directional moves. It does not make you risk-free, and it’s worth stating plainly one more time:

• Gamma still bites. Between rehedges, curvature is uncaptured — that’s the entire hedging error. A short-gamma book bleeds on every round trip and gaps can overwhelm any rehedging schedule.
• Vega is untouched. A jump in implied volatility re-prices your option even if the stock hasn’t moved an inch and your delta is dead-zero. Delta hedging does nothing for it.
• Theta keeps ticking. The clock runs whether or not you rehedge.

To neutralise those, you trade other options (gamma-hedging, vega-hedging) — the subject of later lessons. For now, the headline: delta-neutral kills the first-order bet and nothing more.

Putting it together

Delta-neutral hedging is a process, not a position. You set the hedge ratio to zero out delta, then gamma immediately drifts you off — so you rehedge, and rehedge, balancing a broom that never stops tipping. Black–Scholes promises perfect replication only in the fairy tale of continuous, costless rehedging; discretely you carry a hedging error $\approx \tfrac12\Gamma(\Delta S)^2$ net of theta, whose risk falls like $1/\sqrt{N}$. But each rehedge pays the spread, so transaction costs rise with $N$, and the sum — error plus cost — is U-shaped with a clear optimal frequency that cheap trading pushes higher and expensive trading pushes lower. Through all of it, gamma, vega, and theta still leak: neutral to small moves, never neutral to the world.

Next, we stop treating the other Greeks as leaks to tolerate and start neutralising them on purpose — gamma- and vega-hedging with other options, so a book can be flattened against curvature and volatility, not just direction.