What Is Vol-of-Vol and Why It Should Set Your Position Size

Vol-of-vol measures how violently a name's implied volatility bounces around its own level. It does not tell you whether to take a trade; it tells you what the trade will feel like, which is a position-sizing input most screens ignore.

Two stocks can have identical implied vols, identical implied-versus-realized ratios, identical IV ranks, and utterly different personalities. One name's vol grinds around its level like a commuter train: 41, 43, 42, 44, 42. The other lurches between regimes: 30, 55, 38, 62, 45. Average either series and you get roughly the same number. Hold a short vol position through each and the experience is not remotely the same. The first is a quiet month; the second is a month of marks swinging against you, days where the position looks broken, and a stomach-churning stretch before anything resolves.

That instability of the vol itself is what vol-of-vol measures, and the name means exactly what it says: the volatility of the volatility. It is the third gate in a serious volatility screen, after "am I being paid?" (implied versus realized) and "is this name stretched?" (IV rank), and it answers the question neither of those can: once you are in, how rough is the ride going to be?

The definition: a coefficient of variation

Putting a single number on "how much does this thing bounce around its level" takes two ingredients you already know. The mean: the average level of the name's implied vol over the last month. The standard deviation: the typical distance of daily prints from that average. Then divide:

vol_of_vol = std_of_iv / mean_of_iv

That is the coefficient of variation: standard deviation expressed as a fraction of the mean. A reading of 0.05 means the vol wanders about 5 percent around its level, whatever that level happens to be. A reading of 0.30 means 30 percent swings around the level.

Why not just use the standard deviation?

Because raw standard deviation has a scale problem. A std of 5 vol points is violent for a name whose vol lives at 20 and barely noticeable for a name living at 90. Divide the wobble by the level and the number becomes dimensionless, which is the entire point: a sleepy 20-vol utility and a jumpy 90-vol biotech land on the same yardstick, and only then does ranking an entire universe by ride quality mean anything. This also surfaces something a raw vol level hides completely: high vol-of-vol is not the same thing as high vol. Scan an unstable-decile list on any given day and you will regularly find mid-level vols whose instability is extreme.

What the data says about steady versus unstable names

Here is why the number earns a place next to the ratio and the rank, and the finding surprises most people in one direction. When Alphanume ran the study on full history, taking the fully screened short vol trade (names that were both rich against realized and high in their own IV range) and splitting it by vol-of-vol:

• Steady-vol names produced a tight, predictable distribution of outcomes clustered around a modest win, hitting in roughly 83 percent of cases. Small reversions, few surprises.
• Unstable-vol names captured more on average, their reversions were bigger, and they actually won more often, north of 90 percent. They also came with a far wider, wilder spread of outcomes.

Sit with the second bullet, because the naive read is backwards. High vol-of-vol does not mean "avoid." The big snapbacks live there, precisely because the vol got violently stretched to begin with. What high vol-of-vol tells you is that the path between entry and payoff will be ugly: the running day-to-day valuation of your position swinging against you, a week or two where the trade looks like a mistake, before the reversion resolves. The endpoint is favorable more often; the path is worse the whole way.

A sizing dial, not a veto

The practical use follows directly, and it is a dial with two ends rather than a filter that removes names:

• Low vol-of-vol is where you fish for steady, income-like premium. The reversions are smaller, but outcomes cluster tightly, so you can size the position up and sleep through it.
• High vol-of-vol is where the outsized snapbacks live. They pay more and hit more often, and they will hand you terrifying marks on the way. Take them smaller, so the worst week of the path is survivable.

This is the discipline of reading a backtest honestly, applied before the trade instead of after. A strategy's average hides its path; a per-name vol-of-vol reading is a forward-looking estimate of exactly that path-width, delivered while you can still act on it. Two positions with identical expected edge and very different vol-of-vol readings should not be the same size, and most retail screens have no field that would even let you know the difference exists.

There is a portfolio-level payoff too. A book of five steady-vol premium trades and a book of five unstable-vol trades can show the same expected return on paper while having completely different worst weeks. If your sizing rule only looks at the level of vol, both books look identical at entry. Vol-of-vol is the field that lets the sizing rule see the difference before the market demonstrates it.

How it fits next to the ratio and the rank

It is worth being precise about what each of the three volatility gates knows, because they are frequently confused. The implied-versus-realized ratio compares the level of implied vol to the level of delivered movement: it answers "am I being paid?" IV rank and percentile compare the level of implied vol to the name's own year of levels: they answer "is this print unusual for this name?" Vol-of-vol is different in kind. It is not a comparison of levels at all; it is a property of the recent path of the vol series itself. Two names can agree on every level-based reading and still have completely different personalities, and vol-of-vol is the only one of the three that can tell them apart.

Where the number comes from

The Alphanume Vol-of-Vol dataset computes the coefficient of variation over a trailing 21-observation window, roughly one trading month, per name per day. Each row shows its work: iv_vov alongside the ingredients iv_mean_21 and iv_std_21, the same trio for realized vol, plus daily cross-sectional ranks so "the most unstable decile of the market today" is a single query parameter rather than a computation you maintain. One warm-up quirk to know: the window needs 21 observations, so a ticker's first month in the universe has no vol-of-vol rows at all, and an n_obs_vov field tells you how full the window is. The API documentation covers the full schema.

Run it instead of reading about it

The fastest way to internalize this number is to pull the unstable decile and a famously sleepy mega-cap side by side and look at the gap. That is literally the exercise in the vol-of-vol lesson of Alphanume Learn's Systematic Trading with Market Data course: real Python against the live feed, in the browser, no setup. The lesson closes out the toolkit of the volatility module: paid (the ratio), stretched (the rank), and ride quality (vol-of-vol), before the capstone wires all three into one screen and measures what each filter adds. If you are mapping the broader path, this guide to learning volatility trading sequences the whole arc.

The first module is free, no account needed: start with the ten-minute opening lesson or browse the full syllabus. The rest of the roughly 18-hour course is included with Alphanume Pro (pricing), along with full access to the data platform itself.