Skip to content

Summary

Volatility has two distinct observable forms: historical/realized volatility (HV/RV), the annualized standard deviation of past returns, and implied volatility (IV), the volatility input that forces a pricing model's theoretical value to equal the market price1. IV is therefore not a forecast anyone published — it is a price of options, set by supply and demand for insurance and speculation. The gap between the two is the raw material of volatility trading.

Two Different Objects

Historical (realized) vol Implied vol
What it measures Dispersion of past returns The vol priced into options today
How obtained Computed objectively from a return window Inverted out of market prices via the model
What it reflects What the underlying did Market consensus + hedging/speculative flows
Look-ahead? None (backward-looking) Forward-looking to expiration

Both are quoted in the same units — annualized standard deviation — which is what makes their comparison meaningful. Traders quote IV in vol points, not dollars: "30 bid, 31 offer" means the market trades that option at 30/31% volatility2.

IV Is a Price, Not a Forecast

  • IV is set by supply and demand: hedgers bid it up when protection is feared (earnings, Fed, FDA, takeover rumors); sellers push it down in complacent markets2.
  • IV and price direction are inversely related in equities: stocks fall → put demand rises → IV firms; rallies → protection is sold → IV eases. This is why IV tends to lead realized vol in stress and why the classic pre-earnings divergence appears — HV stalls while IV climbs, and they converge after the event2.
  • Because IV mean-reverts within a per-underlying range, "high" and "low" IV only have meaning relative to that underlying's own history (6–12 months minimum), not against other names2.

Measuring HV: Window Choice Matters

HV is the annualized standard deviation of log returns over a chosen window. The window is a modeling decision, not a fact:

Window Typical use Tradeoff
10-day Reacts fast to regime shifts Noisy; one event dominates
20–30 day Standard trader benchmark (VIX uses 30 calendar days) Balance of responsiveness and stability
60–90 day "Background" vol for regime comparison Slow; blends regimes

Natenberg's volatility-cone idea extends this: plot the distribution of HV over each window across history, so current IV can be judged against the full range of realized outcomes rather than one point estimate1. Deannualization uses the √time rule: 1-day σ ≈ IV/16 (i.e., 32% IV ≈ 2%/day)2.

Volatility as an Asset Class

Once IV is understood as a price, "trading volatility" becomes separable from trading direction. A delta-neutral position isolates the vol exposure: long gamma/vega profits when realized movement exceeds what IV assumed; short gamma/vega profits when it does not1. The persistent tendency of IV to exceed subsequently realized vol — the variance risk premium — is the premium option sellers collect for bearing gap and regime risk; it is compensation, not free money. Systematic harvesting of it belongs to the insurance-business framing of TOMIC and must be evaluated against an explicit IV−RV baseline per the EV contract.

What the Divergence Signals

Passarelli's nine IV/RV patterns reduce to two questions: is the gap between IV and RV persistent or event-driven, and is it justified by an upcoming catalyst?3

  • IV safely above RV, no catalyst on the calendar → short-vol candidate (condors, credit spreads).
  • IV at or below RV and near historical lows → long-vol candidate (long gamma).
  • Means can move: a fundamental change (merger, new business mix) can reset the "normal" vol range, so confirm the historical range still applies.

Links

Source Notes

  1. Natenberg, Option Volatility & Pricing, chs. 4 and 14 (bundle: ../option-volatility-and-pricing-bundle/topics/volatility.md). ↩↩↩

  2. Passarelli, Trading Option Greeks 2e, ch. 3 (bundle: ../trading-option-greeks/topics/volatility.md). ↩↩↩↩↩

  3. Passarelli, ch. 14, volatility charts (bundle: ../trading-option-greeks/topics/volatility-charts.md). ↩