Volatility has two habits that make it forecastable in a way that price is not. It clusters: a big move today makes a big move tomorrow more likely, so calm stretches and stormy stretches come in runs. And it mean-reverts: the runs end, and volatility drifts back toward the stock's own long-run level. A stock that normally realizes 25% and is currently realizing 50% will, in all likelihood, be realizing something closer to 25% a few months from now. Nobody can say when, but the direction is a good bet.
A forecast is not today's IV
Those two habits are what a volatility forecast is built from. A simple forecast blends the most recent realized volatility, which carries the clustering, with the long-run average, which carries the mean reversion, weighting the recent reading more for short horizons and the average more for long ones. ORATS runs a forecast model of this kind and publishes a 20-day forecast for every stock; the Terminal labels it Forecast 20d.
Implied volatility is a different thing. It is what the market is charging, and it includes a risk premium on top of any pure forecast, plus whatever the market knows about events ahead. So there are three numbers to keep apart: what the stock did (HV), what a model expects it to do (the forecast), and what options are priced for (IV). When IV sits well above the forecast, premium is rich even after allowing for the model's view of the future.
Why the front moves and the back does not
Mean reversion also explains a pattern you will see on every stock. A one-week option only has to cover the next week, and next week will probably look a lot like this week, so its implied volatility tracks today's conditions closely. A one-year option covers a year in which today's storm has very likely passed, so its implied volatility stays near the long-run level. When the market panics, short-dated vol jumps and long-dated vol barely moves. When it goes quiet, short-dated vol sinks below long-dated vol.
At-the-money implied volatility by expiration for a stressed stock: 45% at the front, 25% long run. Drag the front-month slider down to 15% for a calm market and the curve slopes upward instead. The one-year end hardly moves either way. The engine draws the term structure only; for the realized-vol cone, use the link.
Plotting implied volatility against time to expiration gives the term structure. Calm markets slope upward, short below long; stressed markets slope downward, short above long. Either way the long end sits near normal, because that is where mean reversion says volatility is headed.
The cone
Mean reversion raises an obvious question: normal compared with what? The volatility cone answers it with the stock's own history. For each window length, 10, 20, 30, 60, 90 and 120 days, take every realized-vol reading over the past year and record where they fell: the median, the middle 50% between the 25th and 75th percentiles, and the outer band from the 10th to the 90th. Plot those against window length and the bands narrow as the window grows, because long windows average away the extremes. That narrowing shape is the cone.
Then draw two lines across it: realized volatility as it stands today at each window, and implied volatility at each tenor. Now every reading has context. Say the 30-day median is 22%, the middle band runs 18% to 27%, and today's 30-day IV is 31%. The market is pricing more movement than this stock realized in three quarters of the past year. That is not automatically a sell; there may be a reason. But it tells you premium is rich against this stock's own record, and mean reversion says the odds favor realized staying nearer the median than the top of the band.
The opposite picture, IV below the median, says options are pricing less movement than this stock usually delivers. Either way, the cone replaces an opinion about whether 31 vol is high with a record of what this stock has actually done.
The maths: a simple mean-reverting forecast
A simple way to write mean reversion: expected vol in h days = LR + (now - LR) × e^(-h / τ). LR is the long-run level, now is today's volatility, h is the horizon in days, and τ (tau) sets how quickly the gap closes; after τ days about 63% of it is gone. The term chart above uses this shape with τ of 60 days.
Worked example: a stock normally at 25% is realizing 50%, with τ of 60 days. In 10 days the expected level is 25 + 25 × e^(-0.17), about 46.2%. In 30 days, 25 + 25 × e^(-0.5), about 40.2%. In 90 days, 25 + 25 × e^(-1.5), about 30.6%. An option covering the next 90 days prices the average over that whole stretch, so its implied volatility sits between today's 50 and the long-run 25. ORATS's own forecast model is more detailed than this, but it is built on the same two habits.
Your turn
Open AAPL. Start on Vol Snapshot for the forecast, then move to IV Intelligence's Cone tab and read the 30d column of the grid. Note whether the implied line sits inside the middle band or above it before you answer the last live question.