A Volatility Index for Any Asset: Regimes, Percentiles and a 30-Day Forecast
2026-10-08
Volatility is how far an asset tends to move, ignoring direction. It is more predictable than price: calm stretches follow calm stretches and storms cluster. The Volatility Index turns that into four numbers for any Time Machine asset: current realized volatility, its percentile against its own history, a regime label and a 30-day forecast.
Bitcoin on 8 October 2026
| Measure | Value |
|---|---|
| 7-day annualised volatility | 26.8% |
| 30-day | 37.6% |
| 90-day | 38.7% |
| Median 30-day since 2017 | 55.1% |
| Percentile of current 30-day | 17th (calm) |
| Forecast for the next 30 days | about 43%, range 30 to 61% |
Bitcoin was in the calm end of its range. Calm is not safe: low volatility often precedes a jump, so the useful reading is "ranges are currently narrow", not "risk is low".
How it works
- Realized volatility is the root mean square of daily log returns over a window, annualised using the asset's own number of trading days a year (about 365 for crypto, about 252 for stocks).
- Regimes come from the percentile of the current 30-day figure: under the 20th is calm, to the 80th normal, to the 95th high, above that extreme.
- Forecast is a HAR-style regression of the next 30 days of realized volatility on the 5-, 30- and 90-day figures, fitted on the asset's own history.
The forecast has to earn its place
The tool backtests the model before showing it. It trains on the first 70% of history, leaves a gap equal to the forecast horizon so no future data leaks in, and compares errors on the rest with simply assuming the current 30-day volatility persists. For Bitcoin since 2017 the regression's log error was 0.279 against 0.304 for persistence over 1,035 test points: about 8% better. If the model does not beat persistence for an asset, the tool says so and falls back to persistence. The band is one standard deviation of past forecast errors, not a guarantee.
What it is for
Volatility tells you how big a move to plan for: position size, stop distance, whether an option or leveraged product is expensive. Whether volatility can improve a trading strategy is a separate question that needs out-of-sample testing; the index is a measurement, not a trade signal. API: /api/lab/volatility?asset=bitcoin&from=2017-01-01.
Pricing and limits
Hindsight runs cost $0.002 in account credits, or come from the monthly plan allowance. A few free anonymous samples are offered each day, and a result someone has already computed is free to view for ten minutes. Spend caps apply to every account.
Numbers come from the BitBank Hindsight Lab on daily closes (Yahoo Finance), retrieved 8 October 2026. They are history, not a forecast, and price only: no dividends, fees or tax.