Gamma Scalping in Practice

Gamma scalping is the repeated delta hedging of a long-gamma option position: buy the underlying after declines as delta falls, and sell after rallies as delta rises. The intuition is simple, but profitability is not. The realized variation captured by hedging must exceed option theta, bid-ask, impact, financing, and losses from implied-volatility changes.

The local accounting identity

For a delta-hedged option under a smooth diffusion,

dPi ≈ Theta dt + 0.5 * Gamma * (dS)^2 + Vega * dIV

Under Black-Scholes, theta and gamma satisfy approximately:

Theta ≈ -0.5 * Gamma * S^2 * IV^2

so expected hedged P&L is positive when the variance actually captured by the hedge exceeds the implied variance paid, before costs. It is a variance comparison, not a comparison of annualized volatilities. A 25% realized-vol estimate is not enough; sampling, jumps, and trading cost determine captured variance.

Component	Long straddle sign	Interpretation
Gamma	positive	benefits from movement
Theta	negative	premium bleed
Vega	positive	benefits if IV rises
Vanna	often material	spot-vol co-movement
Transaction cost	negative	increases with hedge frequency

Choose the option portfolio first

An ATM straddle has high gamma per unit premium but rapid decay. A longer-dated straddle has less gamma, more vega, and lower turnover. A strangle buys cheaper wings but needs a larger move before gamma becomes useful. Strike choice should come from expected realized distribution, liquidity, and desired surface exposure, not simply the highest quoted gamma.

Match the trade to the surface. Buying a straddle at a steeply elevated implied volatility surface before an event requires a genuinely large move or a favorable post-event surface outcome. A delta-hedging backtest that keeps IV constant is a diagnostic toy, not a P&L forecast.

Discrete hedge rules

Continuous hedging is an analytic limit. Actual desks use time, delta, or risk bands. A delta-band rule trades only when residual delta exceeds a threshold:

def rebalance(shares, option_delta, multiplier, band):
    target = -option_delta * multiplier
    if abs(target - shares) >= band:
        return target
    return shares

The optimal band rises with half-spread and impact and falls with gamma and expected volatility. It is not a fixed “rebalance every hour” prescription. Include minimum ticket size, market hours, borrow availability, and a rule for opening gaps. A hedge after a gap can reset delta but cannot harvest the path that was skipped.

Measure captured variance honestly

Use trade-level cash flows. Each hedge changes stock inventory and has an execution price, fee, and estimated impact. Mark the remaining option with a bid/ask-aware surface. Do not label stock trading P&L as gamma P&L unless its timing follows a pre-specified hedge policy.

hedge_pnl += -trade_shares * execution_price - fee
shares += trade_shares
option_mark = surface.price(strike, expiry, spot, side="mid")
book_value = shares * spot + option_quantity * option_mark + hedge_pnl

For research, compare fixed time grids, delta bands, and volatility-scaled bands on the same historical path. Optimize bands only in a training period. Otherwise, the chosen rule overfits microstructure noise and understates live cost.

Major failure modes

Long gamma does not protect against every adverse move. The position can lose after a large gap if the option was bought at even richer implied volatility, and it can lose on a quiet but persistent trend because repeated hedges sell into strength or buy into weakness before the option expires away from the strike. Overnight moves, limit states, halted names, and crypto funding windows make the diffusion approximation particularly unreliable.

Volatility can also fall as realized volatility rises. For example, an event can realize a large but smaller-than-implied move and produce a vega loss greater than harvested gamma. Decompose P&L into realized gamma, theta, vega, vanna, volga, financing, and execution. Reference options Greeks for the definitions, but calculate the attribution by full repricing when the surface moves.

Risk framework

Set a maximum option premium, residual delta limit, overnight policy, and scenario loss limit. Stress a gap before the next hedge, a 5–10 vol-point crush, a skew twist, and doubled spread/impact. Use adjusted data around splits and dividends; otherwise both hedge quantities and realized returns are wrong.

Key takeaways

• Gamma scalping earns realized variance only after paying implied variance and all hedge costs.
• Hedge frequency is a cost-versus-tracking optimization, not a universal schedule.
• Use executed hedge cash flows and surface marks to measure the strategy.
• Gaps and IV crushes can defeat a long-gamma thesis.
• Full P&L attribution prevents vega or execution losses being mislabeled as gamma.