Skip to content

Summary

Short-premium option selling is structurally an insurance business: many small, fairly reliable profits punctuated by rare, large losses. The tail is where the business is won or lost, and the three principles below — negative convexity, discontinuous markets, and defined-risk sizing — determine whether a tail event is a bad quarter or a busted account.

Negative Convexity of Short Premium

  • A short option position is short gamma and short vega: losses accelerate as the market moves (gamma) and as implied volatility rises (vega), and the two compound — a crash simultaneously moves the underlying and reprices vol upward. This is the quadratic term that makes linear models understate VaR for negative-gamma books.3
  • The payoff profile is the mirror of the long-option "card game": sellers collect many small premiums and occasionally pay large claims. The variance risk premium compensates for the left tail — harvesting it means deliberately holding tail risk.2
  • The insurance-company analogy holds at the expected-value level (insurers are profitable over many policies) but hides a capital constraint: an insurer with infinite capital and patience earns the premium; a leveraged trader with margin calls, monthly drawdown limits, and human psychology does not. The tail problem of the analogy is that solvency, not EV, is the binding constraint.

Why Discrete Hedging Fails in Gaps

  • Dynamic delta hedging assumes markets move continuously and liquidly. A gap move offers no reprice window: the hedge cannot be adjusted between the close and the gap, so realized gamma losses exceed anything computed from continuous-time models.2
  • The canonical demonstration is 1987 portfolio insurance: replicating a put by continuous rebalancing performed poorly when gaps and illiquidity broke the replication assumptions exactly when protection was needed.2
  • Correlated failures compound this: hedges based on historical correlation (spreads, cross-market hedges) loosen when correlations converge to 1 in the crash itself (see Diversification and Correlation).
  • Practical rule: assume every short-premium position can open tomorrow 1 gap-day worse without any opportunity to hedge; size (2% rule) and heat limits must survive that assumption.1

Defined-Risk Structures as the Tail Primitive

TOMIC's answer to unbounded tails is structural, not managerial:1

Structure Tail exposure Sizing implication
Naked short put / call Unbounded (call) / to zero (put) Cannot be sized by the 2% rule; excluded from the business
Credit spread / iron condor Capped at wing width Max loss known at entry; the sizing primitive
Long OTM puts / VIX calls ("units") Long the tail; convex payoff Allocate 5–10% of capital as portfolio insurance
Cash Zero market risk A position: skip trading when the environment is wrong
  • Defined risk converts an unquantifiable tail into a bounded, budgetable number: max loss enters Position Sizing arithmetic directly and portfolio heat becomes summable.
  • The "units" sleeve exploits the same convexity that hurts sellers: cheap OTM options with delta < 5 have little gamma/vega until a crash, then gain far more than models predict ("snowball effect").1
  • Take off dying shorts at "card game value" (the residual $0.10–$0.25 the model cannot justify) to free capital and remove the remaining tail before expiry risk concentrates.1

Operating Principles

  1. Never own an unbounded tail you cannot price: defined-risk structures or explicit unit hedges cover every short.1
  2. Assume gaps: back-test and stress-test with overnight gaps and correlation = 1; VaR percentiles alone will understate the short-gamma tail.3
  3. Budget for the tail before selling the head: the units sleeve and the 6% monthly circuit breaker are paid for out of the premium stream, not negotiated during the crash.1
  4. Hedges exchange risk, not eliminate it: collars give up upside; dynamic hedges assume continuity. Choose the risk you are most able to carry.2

Links

Source Notes

  1. TOMIC bundle, topic Risk Management (units, defined-risk sizing, card game value, cash as a position). ↩↩↩↩↩↩

  2. Natenberg bundle, topic Hedging with Options (dynamic hedging assumptions, 1987 portfolio insurance). ↩↩↩↩

  3. Hull bundle, topic Value at Risk (quadratic model, negative-gamma VaR understatement, 22.3-SD moves). ↩↩