Volatility is not a single random variable that draws independently from a fixed distribution day after day. It clusters. Calm periods produce more calm periods; stress produces more stress.

Periods of transition between the two are short, sharp, and consequential. A trader who sizes positions for the 'average' month may have positions that are persistently too small in calm regimes (forfeiting yield) and persistently too large in stress regimes (catastrophic risk). The professional discipline is to classify the current regime explicitly, size for what that regime actually contains rather than the long-run average, and watch transition signals carefully.

The regime-classification framework is not new — Engle's 1982 ARCH and Bollerslev's 1986 GARCH papers formalized the statistical evidence of vol clustering in financial returns, and three decades of subsequent research have refined the toolkit. What has changed in the past decade is the speed and severity of transitions: Volmageddon February 2018, COVID March 2020, the SVB-Credit Suisse banking stress March 2023, and the yen-carry-unwind August 2024 each compressed regime transitions that historically took weeks into hours or days.