Position sizing for option sellers

Size by max loss, not by contracts

One contract can be small or enormous. A cash-secured put risks the strike times 100 less the premium. A credit spread risks the width minus the credit. A position is small only if its maximum loss is small relative to the account, and that is the number to size on.

The payoff shape makes this matter more for sellers than buyers. In our records a winning trade closed early banked about half its credit, while a defensive buyback cost about three times the credit, so one loss erased several wins. A spread can carry a 7 to 1 ratio of maximum loss to credit. A book with that shape needs a high win rate, and it needs each loss to be survivable.

Count factors, not tickers

Positions that move together are one position. Across the tickers in our scanned set, average daily correlation was 0.46, about two independent bets a week. QQQ and its leveraged wrappers moved as one. The useful limit is maximum loss per factor across everything you hold, with any factor that already has a threatened short closed to new premium until it resolves.

The guide on selling options on leveraged ETFs covers why a different ticker on the same index is not a different bet.

What Kelly says, and why it overshoots

The Kelly criterion gives the bet size that maximizes long-run growth for a known distribution. On the monthly returns of a replayed put credit spread program, empirical Kelly came out at about 70% of equity at risk, and half-Kelly at 35%. Both sat where the odds of a 20% drawdown within a year were around 38%.

Kelly assumes you know the distribution. A short premium book's distribution changes with the regime, and the losses cluster. When we mixed bear days into the cash-secured put distribution at their four-year share, the growth-optimal size dropped sharply and the growth at that size nearly vanished. Kelly is a ceiling, not a target.

What share of equity at risk survived

A block bootstrap of 129 real months of the put spread's returns gave these one-year odds of a 20% drawdown: none at 5% or 10% of equity at risk, about 3% at 15%, about 15% at 20%, and 38% at 30%. The median year rose roughly in proportion, from about +4% at 5% to about +25% at 30%.

Our separate study of cash-secured puts put numbers on oversizing: at collateral of 3.4 to 4.5 times equity, the one-year odds of a 20% drawdown were 21% to 31%. At that size a bad year is a one-in-four event. A drawdown cap you can live with, not Kelly, is what should set the size.

Practical checks

Before adding a position, write down its maximum loss, the factor it belongs to, the factor's total maximum loss after the add, and the share of equity that represents. If a second uncorrelated factor is available it raises the growth rate more than adding size to the first; in our replays a second low-correlation underlying cut the drawdown odds at a given size more than any entry filter did.