Volatility Cones: Is This Level Actually Unusual?

Take realised volatility over many overlapping windows — a week, a month, a quarter — and plot the percentiles of each. The result is a cone: wide at short horizons where volatility varies enormously, narrower further out where averaging smooths it.

Drop today’s readings onto it and you can see immediately whether the current level is ordinary or genuinely extreme, and at which horizon.

Why one number cannot do this

A statement like “30-day volatility is 60” is unreadable without context. Sixty is unremarkable for one asset and a historic extreme for another. The cone supplies the context in the same picture, which is the same discipline demanded in funding extremes: an extreme is a claim about a distribution, so show the distribution.

The horizon disagreement is the signal

The genuinely useful reading is when horizons disagree. Short-window volatility at a high percentile while the quarterly window sits mid-range says something specific: the recent tape is unusual against a background that is not. That is a fresh disturbance rather than a sustained regime.

The reverse — long windows elevated while the short window is calm — is a market that has been volatile and is currently resting. Those are different situations that a single volatility number renders identically.

Building one on crypto data

Two adjustments. There are no non-trading days, so the annualisation factor uses every calendar day rather than an equity trading-day count — the error flagged in implied versus realised, and it is a fixed multiple wrong, permanently, if you get it wrong.

And the history is short. A cone needs enough non-overlapping periods for its percentiles to mean anything, so a coin listed recently cannot support one at long horizons. Drawing it anyway produces a cone whose upper bound is simply the largest thing that has happened yet.

Reading it against the options market

The cone describes what the asset has done. The volatility surface describes what the options market charges for what it might do. Putting implied onto the realised cone is the cleanest single view of whether options are expensive, at each horizon separately.