Historical Volatility: A Backward Look
The historical volatility indicator plots the standard deviation of past returns, usually annualised, as a line beneath the chart. It measures what has already happened, which is what separates it from implied volatility - the market's forward estimate priced into options.
How it works
The indicator computes the standard deviation of returns over a rolling window and plots it as a line. Each point is a summary of the bars behind it.
It is usually annualised. A daily deviation multiplied by the square root of 252 — the approximate number of trading days in a year — puts it on the same scale as an options quote.
On this site’s shared 576-bar history the bar-to-bar deviation of 0.347% annualises to about 5.5%. That is a low figure by equity standards, and the series is a generated one rather than a market.
Backward, not forward
The line lags by construction. A quiet reading after a violent week means the violence has left the window, not that conditions have calmed.
Implied volatility is the forward estimate. It is derived from option prices and represents what the market is charging for future movement, which is a different quantity with a different use.
The one genuine forecasting property is clustering. Volatility is persistent — a high reading is more often followed by another high reading than by a low one, which is why a backward measure has any predictive value at all.
The window length is the only real setting. A 10-bar reading reacts within days and is noisy; a 100-bar one is stable and slow. Neither is correct — they answer different questions, and using one where the other belongs is the most common error with the tool.
In practice
Nothing in the calculation sees volume. A large move on almost no participation raises the reading exactly as much as one on heavy trading.
Stability and lag are the same dial. Every step that makes the line calmer also makes it slower, and there is no setting that avoids the trade-off.
A gap enters as one very large return and, because deviations are squared, it can dominate a short window for as long as it stays inside it.
The honest use is sizing. The reading tells you how wide a stop needs to be in current conditions; where to put it is a structural question the indicator cannot answer.
Low volatility makes costs proportionally worse, not better. The round trip is 2% of a median bar’s range on this history and a larger share of a quiet one, so quiet conditions are the expensive ones to trade actively.
It is a description, not a participant. Nothing in the order book responds to the line, which is why it belongs in the sizing decision rather than the entry one.
Reading it as a percentile
A raw volatility number means nothing without a reference. 5.5% annualised is high for a government bond and very low for a small-cap stock, so the reading only becomes usable once it is placed against the instrument’s own history.
The practical version is a percentile rank over the past year. A reading in the bottom decile of its own range says positions can be larger in units and stops tighter in points; a reading in the top decile says the reverse. That single transformation turns an uninterpretable number into a sizing instruction, and it costs one extra line of calculation.
There is a comparison worth making that most platforms will not draw for you: the two readings side by side. Where implied volatility sits well above the historical figure, options are expensive relative to what the instrument has actually been doing; where it sits below, they are cheap by the same standard.
That spread is the input to every volatility-based options decision, and it is also a useful sentiment reading for someone who never trades an option. A large premium says the market expects something the recent past does not contain. It is not a forecast either — the expectation can be wrong, and frequently is — but it is one of the few numbers available that reflects what other participants are actually paying for.
What historical volatility is not
It is not a forecast. It measures a window that has closed.
It is not implied volatility. That is priced, not measured.
It is not direction. High readings occur in both.
And it is not risk. Variability and losing money are different things.
When it fails
A range produces the indicator’s most dangerous reading. Volatility compresses, the line reaches a low, position sizes computed from it reach their largest, and the breakout that follows arrives against the biggest position the method will ever hold.
The second failure is treating a low reading as safety. It is a statement about the past fortnight and frequently the opposite of a forecast.
A third is a window shorter than the events you care about. A 10-bar reading forgets a crash in a fortnight.
A fourth is comparing raw readings between instruments. Only a percentile rank is comparable.
And a fifth is expecting a signal. It has no cross, no threshold and no direction, and every attempt to manufacture one adds a parameter without adding information.
The original data
On this site’s shared 576-bar history the 575 bar-to-bar returns have a standard deviation of 0.347%,
annualising to about 5.5% at 252 periods. The largest single moves were +1.065% and -1.194%, and 8 moves
exceeded three deviations where a normal distribution implies fewer than two. The figures are in
research/series-measurements.json, produced by site/measure_series.py.
Those eight outliers are the reason to size against the tail rather than the average. A method built around a 0.347% typical move meets a 1.194% one several times a year, and a position sized for the typical day is three and a half times too large on that day. Set size from a bad case rather than a normal one, and the low-volatility trap that catches most of the damage in this tool simply does not arise.
Related
Standard deviation is the underlying calculation and its assumptions. Implied volatility is the forward-looking counterpart. And Average true range is the range-based alternative that needs no return series at all.
I use it for one thing: sizing. A reading in the bottom quarter of its own range tells me my usual stop distance is currently too wide in points and my usual size is too small, and a reading in the top quarter tells me the opposite. It has never once told me which way anything was going.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.