Home / Market Regimes / Mean Reversion in Volatility
Market Regimes Intermediate The pull of volatility toward a long-run level Forward-looking

Mean Reversion in Volatility

Mean reversion tells you a high reading will probably fall. It does not tell you when, and margin calls do not wait for the long run.

Quick answer: Mean reversion in volatility is the persistent tendency of volatility to be pulled back toward a long-run level — around 14% annualised for NIFTY — with excursions above the mean that are violent and short-lived and excursions below it that are shallow and long-lasting, an asymmetry that gives the distribution of volatility its long right tail.

In simple words

Volatility does not wander off forever. It is pulled, persistently, back toward a long-run level — for NIFTY that level is roughly 14% annualised. But the pull is not symmetric. When volatility shoots above the mean, the moves are violent and short-lived, because a spike to 40% is a crisis and crises do not last. When volatility sits below the mean, it can stay there for months, drifting quietly, because volatility cannot go below zero and there is a floor to how still a market can be. So the picture is a series of long, calm valleys interrupted by sharp, brief peaks. Suppose India VIX is at 26 after a shock. Mean reversion says it will probably come back toward 14 — and it probably will. What it does not say is when, or whether it climbs to 35 first.

That asymmetry — quick violent peaks, long shallow valleys — is the reason the distribution of volatility has a long right tail. Most of the time volatility is low or normal, but occasionally it is enormous, and the enormous readings pull the tail far to the right while the low readings pile up against the floor at zero. This is not a piece of trivia. It is the direct reason IV Rank and IV Percentile disagree: one measures where today sits between the lowest and highest readings of the past year, and the other measures the fraction of days that were lower than today. When the distribution has a long right tail, those two answers come apart, and knowing why is knowing something real about the shape of volatility itself.

Not to be confused with: Mean reversion in prices, which is a much weaker and more contested claim. A price series is close to a random walk — knowing it is high tells you little about where it goes next — whereas volatility genuinely and reliably mean-reverts. Confusing the two leads people to expect volatility to behave like a price, or worse, to trade a high volatility reading with the same confidence they would never apply to a high price. The mean reversion is real for volatility; the timing is not deliverable for either.

The pull toward the long-run mean

Violent short peaks, long shallow valleys

A simulated volatility path mean-reverting toward roughly 14%, with the long-run level drawn as a horizontal line.

the gravitational band5%10%15%20%25%30%35%40%0d100d200d300d400dlong-run mean 16.1%spikes are violent and brief……troughs are shallow and longTrading dayImplied volatility
The picture proves the asymmetry that the average hides. Excursions above the line are tall and brief; excursions below it are shallow and long. The mean is a real attractor — the path always comes back to the line — but it comes back on no schedule, and it frequently overshoots on the way there. A trader who reads 'reverts to the mean' as 'reverts soon' has read the line and missed the horizontal axis.

Professional explanation

The mean is a real attractor, and it is around 14% for NIFTY

Unlike a price, which is close to a random walk and has no level it is drawn back to, volatility has a long-run mean and is genuinely pulled toward it. For NIFTY that long-run level is roughly 14% annualised, and India VIX spends most of its life in the 11%-to-16% band around it. The attraction is not a metaphor: statistically, a high reading is followed on average by a lower one and a low reading by a higher one, reliably enough that it is one of the few forecastable features of financial markets. This is the foundation on which volatility trading, volatility forecasting and the entire volatility-index construction rest. But 'the mean is an attractor' is a statement about the average of many paths, not about any single path, and a single path is all a trader ever actually holds. The average reverts. Your position lives on one draw.

The asymmetry: violent peaks, shallow valleys

Excursions above the mean and excursions below it are not shaped alike. Above the mean, volatility rises violently and falls back relatively quickly, because a reading of 30% or 40% corresponds to a crisis, and crises are intense but do not persist for months — the same escalator-and-stairs asymmetry that governs a dislocation still resolves faster than a low regime unwinds. Below the mean, volatility drifts down shallowly and can stay depressed for a long time, because there is a hard floor at zero and a soft floor set by how still a real market can actually be — prices never stop moving entirely, so realised volatility cannot collapse to nothing. The result is a series of long, quiet valleys punctuated by sharp, brief peaks. This asymmetry is not a detail; it is the reason volatility's distribution is right-skewed, and right-skew is the reason two reasonable ways of ranking a volatility reading against its own history can give materially different answers.

The mathematics: Ornstein–Uhlenbeck and CIR

The standard continuous-time description of mean reversion is the Ornstein–Uhlenbeck process, dσ = κ(θ − σ)dt + ξ√σ dW in its square-root (Cox–Ingersoll–Ross) form. Read it as a tug of war. The drift term κ(θ − σ)dt is the pull: whenever σ is above the long-run mean θ, the bracket is negative and the process is pushed down; whenever σ is below θ, the bracket is positive and it is pushed up; and κ sets how hard. The diffusion term ξ√σ dW is the noise that keeps knocking it away from θ, where ξ is the volatility of volatility and dW is the random shock. The √σ factor is what keeps volatility from going negative — as σ approaches zero the noise shrinks, so the process cannot cross the floor. The speed κ has a physical meaning through the half-life ln(2)/κ, the expected time for an excursion to close half its distance to the mean. For volatility that half-life is measured in weeks; for a price it is not measurable at all, because a price has no mean to revert to, so ln(2)/κ is undefined for it. That single contrast — volatility has a finite reversion half-life and a price does not — is why one is forecastable and the other is not.

The critical caveat: reversion is not a clock

This is the most important paragraph on the page, and it is the one most often skipped. Mean reversion tells you that a high volatility reading will probably fall. It does not tell you when, and it does not stop the reading from doubling first. The half-life is an expectation over many paths, not a countdown on yours: an excursion with a three-week half-life can still spend two months elevated, or spike another 50% before it turns, and 'probably lower eventually' is no comfort to a position that is stopped out or margin-called in the meantime. Margin calls do not wait for the long run. This is, by a wide margin, the single most misused idea in retail volatility trading: 'volatility always comes back down, so sell it up here' treats a statistical tendency as a timing signal, and being early with a short-volatility position is financially identical to being wrong, because you can be liquidated before the reversion you correctly predicted ever arrives. The mean reversion can be completely real and the trade can still ruin you.

Why volatility futures price toward the mean, not toward spot

Mean reversion is the origin of the entire term structure of volatility futures. Because a high spot volatility is expected to fall and a low spot volatility is expected to rise, a futures contract on volatility is not priced off today's spot — it is priced toward the long-run mean, adjusted for how much reversion is expected to happen before the contract expires. When spot volatility is low, the futures curve slopes upward (contango), pricing the expected rise toward the mean; when spot volatility is high, the curve slopes downward (backwardation), pricing the expected fall. This is the mechanical source of roll yield: a long volatility-futures position in a calm, contango market pays away the gap between the low spot and the higher futures as time passes and the contract rolls down toward spot. The contango-and-roll-yield structure that dominates volatility-linked products is not an accident of those products — it is mean reversion, expressed as a price. Anyone trading the roll is, whether they say so or not, trading the speed and direction of mean reversion.

Why IV Rank and IV Percentile disagree

The right-skew that mean reversion produces is exactly why the two most common ways of contextualising a volatility reading give different numbers. IV Rank places today between the lowest and highest readings of the past year — it is a position on a line whose top end is set by a single crisis peak. IV Percentile counts the fraction of days that were lower than today — it ignores how far the extremes reach and only cares about ordering. When the distribution is right-skewed, a rare spike stretches the top of the IV Rank scale far above the crowd of ordinary readings, so a moderately elevated day can show a low IV Rank (it is far from the crisis high) while simultaneously showing a high IV Percentile (it is higher than most ordinary days). Neither is wrong; they answer different questions about the same skewed distribution. Understanding that the disagreement comes from the long right tail — which comes from the asymmetry of mean reversion — turns a confusing pair of indicators into two complementary readings of one underlying shape.

Why the distribution has a long right tail

The histogram of volatility readings implied by mean reversion against a floor at zero.

00.10.20.30.40.5-4σ-3σ-2σ-1σ0σ1σ2σ3σ4σthe tails thatmattera fat-tailed market has MORE quiet daysand MORE extreme days than the bell allowsStandardised daily return (in standard deviations)Probability densityNormal (Black–Scholes assumption)Fat-tailed (real returns)
Because volatility cannot fall below zero and can spike almost without limit, the readings pile up on the left against the floor and stretch far to the right in a long thin tail. This is the shape that makes IV Rank and IV Percentile disagree, and it is why the average volatility reading is higher than the most common one — the rare crises drag the mean above the mode.

Formula

The mean-reverting (Ornstein–Uhlenbeck / CIR) process for volatility

dσ = κ(θ − σ) dt + ξ√σ dW

A tug of war between a pull and a push. The drift κ(θ − σ)dt pulls σ toward the long-run mean θ — negative when σ is above θ, positive when below — with κ setting the strength. The diffusion ξ√σ dW is the random noise that keeps knocking it away. The √σ factor shrinks the noise as σ nears zero, which is what keeps volatility from going negative. This square-root form is the Cox–Ingersoll–Ross process; drop the √σ and it is the plain Ornstein–Uhlenbeck process.

  • dσThe infinitesimal change in volatility over an instant dt.
  • σThe current level of volatility (annualised).
  • κSpeed of reversion — how hard volatility is pulled toward the mean. Larger κ means faster reversion and a shorter half-life.
  • θThe long-run mean level toward which volatility reverts — roughly 14% annualised for NIFTY.
  • dtAn infinitesimal increment of time.
  • ξVolatility of volatility — the size of the random shocks that push σ around. Often written as the Greek letter xi.
  • dWThe increment of a Wiener process (standard Brownian motion) — the random, mean-zero shock over dt.

The half-life of a volatility excursion

t½ = ln(2) / κ

The expected time for an excursion to close half its distance back to the mean θ. For volatility this half-life is measured in weeks; for a price it is not measurable at all, because a price has no mean to revert to and κ is effectively zero for it. That contrast is why a volatility forecast is a reasonable thing to attempt and a price forecast usually is not.

How to use mean reversion without being ruined by it

  1. Anchor on the long-run mean, not on spot. Ask where volatility sits relative to its long-run level — around 14% for NIFTY — rather than relative to yesterday, because the reversion is toward the mean and yesterday is not the mean.
  2. Estimate the direction of the pull, and accept you cannot estimate the timing. If a reading is well above the mean, reversion says the expected next move is down; the half-life tells you the average speed but says nothing about your particular path.
  3. Size for the excursion continuing, not for the reversion arriving. Before putting on any position that profits from reversion, ask how far the reading can move against you first and whether your margin and stop survive a further spike.
  4. Read IV Rank and IV Percentile together, knowing they disagree because the distribution is right-skewed. A low IV Rank with a high IV Percentile is not a contradiction; it means today is far from the crisis high but above most ordinary days.
  5. If you trade volatility futures or volatility-linked products, recognise the curve as mean reversion priced. Contango when spot is low, backwardation when spot is high, and the roll yield you earn or pay is the market's estimate of the reversion between now and expiry.
  6. Never convert 'it will probably revert' into 'therefore I should be short here'. Separate the correct forecast from the position: a true statement about the average path is not a licence to hold a leveraged position through the one path you actually get.

Practical example

NIFTY worked example

Take India VIX at 26 — stressed, well above its long-run mean of about 14. Suppose the excursion has a reversion speed κ of 5 per year. The half-life is ln(2) ÷ 5 = 0.139 years, which is about 51 calendar days, or roughly seven weeks. So the expected path is that half the gap between 26 and 14 — that is, six points — closes over about seven weeks, bringing the reading to around 20, with another seven weeks expected to halve the remaining gap to roughly 17, and so on. Interpret this carefully. The forecast is genuinely useful: the central expectation really is downward, and over a quarter the reading is more likely to be near 17 than near 26. But now read the same number as a trader. Seven weeks is the half-life of the average of many paths; your single path can easily spend that seven weeks at 30 or spike to 35 before it turns, and a short-volatility position put on at 26 'because it will revert' can be margin-called at 35 long before the reversion you correctly forecast ever arrives. The mean reversion is real, the arithmetic is right, and the trade can still end you. That gap — between a correct forecast and a survivable position — is the entire lesson.

BANKNIFTY worked example

BANKNIFTY has a higher long-run mean than NIFTY — call it roughly 17% against NIFTY's 14% — because it is a concentrated single-sector index that genuinely realises more movement. The mean-reversion lesson it teaches is about the mean itself, not just the pull. Suppose BANKNIFTY implied volatility is at 22%. Against NIFTY's mean of 14% that looks stretched and ready to revert hard; against BANKNIFTY's own mean of 17% it is only modestly elevated and may drift back slowly, or not much at all. A trader who imports NIFTY's long-run mean onto BANKNIFTY will systematically over-estimate how far BANKNIFTY 'should' fall and will keep selling a volatility that is closer to its own normal than it looks. The reversion target is underlying-specific: each index reverts to its own θ, and using the wrong θ turns a correct idea into a standing source of error. Comparing raw levels across the two indices tells you little; comparing each against its own long-run mean tells you where the pull actually points.

Lot sizes used above (NIFTY 75, BANKNIFTY 30) are those in force at the time of writing; NSE revises them periodically. Figures exclude brokerage, STT, exchange charges, stamp duty and GST. Examples are teaching scenarios built on round numbers — they are not historical quotes, not backtests and not trade calls.

Risk note. Mean reversion is the justification most often given for the single most misused trade in retail volatility: shorting a high volatility reading 'because it always comes back down'. The tendency is real, but it is a statement about the average of many paths, not about the one path your position lives on. A high reading can double before it reverts, and being early with a short-volatility position is financially identical to being wrong, because margin calls and stops do not wait for the long run — you can be liquidated before the reversion you correctly predicted arrives. Short-volatility positions carry theoretically unlimited loss. A correct forecast about the mean does not make the position between here and the mean survivable, and treating reversion as a timing signal is how confident traders are ruined by a call that later turns out to have been right.

Advantages & limitations

What it is good for

  • Volatility mean reversion is one of the genuinely forecastable features of financial markets. Unlike price, a high volatility reading really does carry information that the next reading is more likely to be lower, which is what makes volatility forecasting a defensible activity at all.
  • It gives volatility a stable anchor — the long-run mean — against which any reading can be judged. 'Is this high?' becomes a well-posed question because there is a level, around 14% for NIFTY, that the number is being high relative to.
  • It explains the entire term structure of volatility futures in one idea: contango when spot is low, backwardation when spot is high, and roll yield as the price of the expected reversion. A trader who understands mean reversion understands where the curve's shape comes from rather than memorising it.
  • It makes the disagreement between IV Rank and IV Percentile intelligible rather than mysterious, because both fall out of the right-skewed distribution that mean reversion's asymmetry produces — turning two confusing indicators into two complementary views of one shape.
  • The half-life ln(2)/κ gives a concrete, estimable sense of speed. Even though it does not deliver timing for a single path, it lets a risk manager reason about how long an elevated regime is expected, on average, to persist.

Where it breaks down

  • It gives direction, never timing. The half-life is an expectation over many paths, so mean reversion cannot tell you when a reading will fall, and the interval before it does is exactly where a leveraged position is at risk. A tool that is silent on timing is dangerous precisely for the trades that need timing most.
  • The long-run mean is not constant and not the same across underlyings. NIFTY's roughly 14% is a convention drawn from recent regimes, BANKNIFTY's is higher, and both can shift as market structure changes — so a θ imported from the wrong index or the wrong era produces systematically biased conclusions.
  • The parameters κ and θ are estimated from history, and history that spans a regime change describes no single regime cleanly. An estimate of reversion speed fitted across a period that contained both a calm stretch and a crisis is an average of two different behaviours and may match neither the market you are in now.
  • The square-root process keeps volatility positive but still assumes a smooth, continuous pull, when real volatility jumps. A shock can move volatility discontinuously in a way no reversion parameter anticipated, and the model quietly assumes the path is continuous when the dangerous moments are exactly the discontinuous ones.
  • Reversion says nothing about direction of the underlying. A volatility reading reverting toward the mean is equally consistent with the index rising or falling, so a mean-reversion view on volatility is not, and must never be read as, a view on where prices go.

Common mistakes

  • Treating 'it will revert' as 'it will revert soon'. Mean reversion is a statement about the average path, not a clock, and a short-volatility position put on because a reading is high can be stopped out or margin-called while it doubles first — being early is financially identical to being wrong.
  • Importing one underlying's long-run mean onto another. Judging BANKNIFTY's volatility against NIFTY's roughly 14% mean makes BANKNIFTY look permanently stretched, leading to a standing bias toward selling a volatility that is near its own normal of about 17%.
  • Confusing mean reversion in volatility with mean reversion in price. Volatility genuinely reverts; price is close to a random walk. Trading a high price with the confidence appropriate to a high volatility reading, or vice versa, mixes up a reliable tendency with a contested one.
  • Estimating κ and θ across a window that spans a regime change. The resulting reversion speed and mean are averages of two different worlds, and using them to size a position in today's regime calibrates you to a market that never existed.
  • Reading IV Rank and IV Percentile as if they should agree, and distrusting the data when they do not. They diverge because the distribution is right-skewed; the divergence is information about where today sits, not an error to be reconciled away.
  • Sizing a reversion trade to the expected outcome instead of to the adverse path. Planning for the reading to fall to the mean, rather than for it to spike further first, is how a correct forecast becomes an account-ending position — the loss happens on the road to the destination you got right.

Professional usage

Volatility desks build mean reversion into their forecasts and their curve trades explicitly. A forecasting model for realised volatility is, at its core, a mean-reverting model — a GARCH or a stochastic-volatility specification whose long-run variance is exactly the θ of this page — and the desk's edge is a better estimate of κ and θ than the market's, applied with continuous delta-hedging so the profit and loss depends on the volatility forecast rather than on direction. On the curve side, a relative-value trader reads the volatility-futures term structure as mean reversion priced and takes positions on whether the market's implied reversion speed is too fast or too slow: selling an overly steep contango when spot volatility is low, or fading a backwardation that prices reversion the trader thinks will be slower than the curve assumes. In both cases the discipline is the same one this page insists on — the reversion is the thesis, and the position is sized so that being early does not force an exit before the thesis pays.

Risk managers use the mean as the tether for stress scenarios and for sizing tail hedges. Because volatility reverts, a stress test does not have to assume a crisis lasts forever — it can model an elevated regime decaying over a plausible half-life — but a careful risk manager also refuses to let the reversion assumption soften the near-term scenario, because the loss that matters is taken on the excursion, before any reversion helps. And product-structuring desks that build volatility-linked notes price the roll yield of the underlying futures directly from the reversion parameters, because the carry those products earn or pay in calm markets is nothing more than mean reversion converted into a cash flow. Across all these desks the through-line is that mean reversion is treated as a real and usable regularity whose one fatal misuse — as a timing signal for a leveraged short — is guarded against by design rather than by hope.

Key takeaways

  • Volatility is genuinely and reliably pulled toward a long-run mean — around 14% for NIFTY — unlike price, which is close to a random walk. That is why a volatility forecast is a reasonable thing to attempt and a price forecast usually is not.
  • The asymmetry is the point: excursions above the mean are violent and short-lived, excursions below are shallow and long-lasting, because volatility cannot go below zero. That asymmetry gives the distribution its long right tail, which is why IV Rank and IV Percentile disagree.
  • The mean-reverting process dσ = κ(θ − σ)dt + ξ√σ dW has a half-life ln(2)/κ measured in weeks. A price has no such half-life at all, which is the whole reason the two behave differently.
  • Mean reversion tells you a high reading will probably fall. It does not tell you when, and it does not stop the reading from doubling first. Margin calls do not wait for the long run — this is the single most misused idea in retail volatility trading.
  • Because volatility reverts, volatility futures price toward the long-run mean rather than toward spot, which is the origin of the entire contango, backwardation and roll-yield structure.

Mean reversion in volatility is the rare thing in markets that is both real and forecastable, and precisely because of that it is the most dangerous idea to hold carelessly. The pull toward the long-run mean is genuine: a high reading really will probably fall, a low one really will probably rise, and the whole apparatus of volatility forecasting and volatility-futures pricing rests on it. But the pull acts on the average of many paths, and a trader holds exactly one. The reading can double before it reverts, and being early with a leveraged short is financially identical to being wrong, because you can be liquidated before the reversion you correctly predicted ever arrives. The sentence a marketing department would cut is this: a correct forecast is not a survivable position, and the surest way to be destroyed by volatility is to be right about where it is going and wrong about how much it can hurt you on the way there. Use the mean as an anchor and a curve-pricing engine; never use it as a clock.

Frequently asked questions

What is mean reversion in volatility?
Mean reversion in volatility is the persistent tendency of volatility to be pulled back toward a long-run level — around 14% for NIFTY. High readings tend to be followed by lower ones and low readings by higher ones, reliably enough that it is one of the few genuinely forecastable features of markets.
What is the long-run mean volatility for NIFTY?
Roughly 14% annualised, with India VIX spending most of its life in the 11%-to-16% band around it. This is a convention drawn from recent regimes rather than a fixed constant, and it differs across underlyings — BANKNIFTY's long-run mean is higher because it realises more movement.
Why is mean reversion in volatility asymmetric?
Because excursions above the mean are violent and short-lived — a spike to 40% is a crisis, and crises do not persist — while excursions below the mean are shallow and long-lasting, since volatility cannot go below zero and prices never stop moving entirely. That asymmetry gives the distribution its long right tail.
What is the half-life of a volatility excursion?
The half-life is ln(2)/κ, the expected time for an excursion to close half its distance back to the mean. For volatility it is measured in weeks; if the reversion speed κ is 5 per year, the half-life is about 51 days. It is an average over many paths, not a countdown on any single one.
Does mean reversion tell me when volatility will fall?
No, and this is the critical caveat. Mean reversion gives direction, never timing. A high reading will probably fall, but it can double first, and the half-life describes the average of many paths, not the one your position holds. Margin calls do not wait for the long run.
Why is mean reversion the most misused idea in retail volatility trading?
Because 'volatility always comes back down, so sell it up here' treats a statistical tendency as a timing signal. Being early with a short-volatility position is financially identical to being wrong — you can be liquidated before the reversion you correctly predicted ever arrives. The forecast can be right and the trade still ruinous.
What is the Ornstein–Uhlenbeck process?
It is the standard continuous-time model of mean reversion: dσ = κ(θ − σ)dt + ξ dW, where a drift term pulls σ toward the mean θ at speed κ and a noise term ξ dW knocks it around. The square-root (CIR) variant multiplies the noise by √σ to keep volatility from going negative.
What does the CIR square-root form add?
The √σ factor in ξ√σ dW shrinks the random shocks as volatility approaches zero, which prevents the process from crossing into negative volatility. It is the Cox–Ingersoll–Ross form; without the √σ it is the plain Ornstein–Uhlenbeck process, which can in principle go negative.
Why does a price not mean-revert the way volatility does?
Because a price is close to a random walk with no level it is drawn back to, so its reversion speed is effectively zero and its half-life ln(2)/κ is undefined. Volatility has a finite half-life measured in weeks. That contrast is why volatility is forecastable and price direction usually is not.
How does mean reversion explain volatility futures pricing?
Because a high spot volatility is expected to fall and a low one to rise, volatility futures are priced toward the long-run mean rather than toward spot. Low spot gives an upward-sloping (contango) curve, high spot gives a downward-sloping (backwardation) curve, and the roll yield is the price of the expected reversion.
What is roll yield and where does it come from?
Roll yield is the gain or loss a futures position earns as the contract rolls toward spot over time. In a contango volatility market a long position pays away the gap between low spot and higher futures — that gap exists only because mean reversion prices the expected rise toward the mean. Roll yield is mean reversion turned into a cash flow.
Why do IV Rank and IV Percentile disagree?
Because the distribution of volatility is right-skewed, and the two measure different things. IV Rank places today between the year's low and high — a scale whose top is set by one crisis peak — while IV Percentile counts the fraction of days lower than today. On a skewed distribution those answers come apart, and neither is wrong.
Can I use mean reversion to time a short-volatility trade?
You can use it for direction but not for timing, and timing is what a leveraged short needs most. The reversion is an average over paths; your single path can spike far further before it turns. Size any such position for the excursion continuing, not for the reversion arriving. Short-volatility positions carry theoretically unlimited loss.
Is the long-run mean the same for every index?
No. NIFTY's is around 14%, BANKNIFTY's is higher — roughly 17% — because it is a concentrated single-sector index that realises more movement. Judging one index's volatility against another's mean is a standing source of error; each reverts to its own θ.
Does mean reversion say anything about market direction?
Nothing at all. Volatility reverting toward its mean is equally consistent with the index rising or falling, so a mean-reversion view on volatility is not a view on price. Reading a reverting volatility as a directional signal is a category error.
Why does the distribution of volatility have a long right tail?
Because volatility cannot fall below zero but can spike almost without limit, low readings pile up against the floor while rare crises stretch the upper tail far to the right. The asymmetry of mean reversion — brief violent peaks, long shallow valleys — is what produces that skew.
What is the volatility of volatility (ξ)?
It is the parameter ξ in the mean-reverting process that sets the size of the random shocks pushing volatility around — how volatile volatility itself is. A higher ξ means wider swings around the mean and a fatter right tail, and it is a traded quantity in its own right through products sensitive to the volatility of volatility.
How is κ estimated, and why is that hard?
The reversion speed κ is estimated from historical volatility data, and it is hard because a sample that spans a regime change mixes two different behaviours. An estimate fitted across both a calm stretch and a crisis is an average that may match neither, so the half-life it implies can be wrong for the regime you are actually in.
Does India VIX mean-revert?
Yes, reliably over long horizons: it clusters, so a high reading tends to be followed by more high readings, and it mean-reverts, so no level persists forever. As with any single index, that tells you direction over time, not when a specific elevated reading will come back down.
If mean reversion is real, why do volatility sellers ever lose?
Because reversion has no clock and no upper bound on the excursion. A seller can be correct that volatility will eventually revert and still be margin-called or stopped out when it doubles first, and in a crisis implied volatility can even trade below realised, so the option sold was cheap relative to the movement. The tendency is real; survival to collect on it is not guaranteed.
How should a risk manager use mean reversion?
As a tether for stress scenarios — modelling an elevated regime decaying over a plausible half-life rather than lasting forever — while refusing to let that reversion assumption soften the near-term loss, because the damage is taken on the excursion before any reversion helps. It anchors the scenario without excusing the tail.
Is mean reversion in volatility the same as volatility clustering?
They are two faces of the same behaviour but distinct claims. Clustering says a high reading tends to be followed by more high readings (persistence); mean reversion says the reading is nonetheless pulled back toward a long-run level over time. Clustering explains why regimes last; mean reversion explains why they eventually end.

Voice search & related questions

Natural-language questions people ask about mean reversion in volatility.

Does volatility always come back down?
Over long horizons it reliably reverts toward a long-run mean near 14% for NIFTY, so a high reading probably falls eventually. But 'eventually' is doing heavy lifting: it can rise a lot further first, and mean reversion never tells you when, which is exactly the information a trade needs.
If volatility is high, should I just sell it?
That is the most dangerous instinct in volatility trading. A high reading can double before it reverts, and being early with a short is the same as being wrong once you are margin-called out of it. Short-volatility positions carry theoretically unlimited loss, and nothing here is advice to put one on.
How long until high volatility goes back to normal?
On average, a number of weeks set by the half-life ln(2)/κ — if the reversion speed is around 5 per year, roughly seven weeks to close half the gap. But that is the average of many paths; your particular path can stay elevated far longer or spike again before turning.
Why is forecasting volatility easier than forecasting price?
Because volatility genuinely reverts to a mean and price is close to a random walk. Volatility has a finite reversion half-life measured in weeks; price has no mean to revert to at all. That single difference is why a volatility forecast is defensible and a price forecast usually is not.
Why do IV Rank and IV Percentile give me different numbers?
Because the distribution of volatility is lopsided, with a long tail of rare high readings. IV Rank measures distance from the year's crisis high; IV Percentile counts how many days were lower than today. On a skewed distribution those disagree, and both are telling you something true.
Why are volatility futures priced differently from spot volatility?
Because mean reversion means today's spot is expected to move toward the long-run mean, so the future is priced toward that mean, not toward spot. When spot is low the curve slopes up; when spot is high it slopes down. The roll yield you earn or pay is that expected reversion, priced.
Can I lose money being right that volatility will fall?
Yes, and it happens constantly. If you short volatility at a high reading and it spikes further before it reverts, you can be liquidated before the fall you correctly predicted arrives. Margin calls do not wait for the long run, so a right forecast and a wrong position size end the same way.
Is BANKNIFTY volatility high just because it is above NIFTY's mean?
No — that is a common trap. BANKNIFTY reverts to its own higher mean, around 17%, because it realises more movement as a concentrated sector index. A reading that looks stretched against NIFTY's 14% may be near normal for BANKNIFTY, so you have to use each index's own long-run mean.

Sources & references

Last reviewed 10 July 2026. Educational content only — not investment advice.

← Previous
Crisis Volatility
Educational content only — not investment advice. Every diagram on this page is generated from the site's own model, using illustrative inputs rather than live quotes. Options and futures carry substantial risk, including loss exceeding your deposit on short-volatility positions. See our Risk Disclosure and SEBI Disclaimer.