Core Volatility Beginner Volatility rescaled to one year

Annualized Volatility

The number everyone quotes, computed with two different calendars nobody mentions.

Quick answer: Annualized volatility is a short-period volatility rescaled to a one-year horizon by multiplying by the square root of the number of periods in a year, because variance grows linearly with time while volatility grows with its square root.

In simple words

Annualized volatility takes a volatility you measured over a short period — a day, a week, a month — and expresses it as if it applied to a whole year, so that every volatility number is on the same scale and can be compared. Suppose NIFTY's daily returns have a standard deviation of about 0.85% — that is a typical day's wobble. To turn that into an annual figure you do not multiply by 252, the number of trading days; you multiply by the square root of 252, which is about 15.87. So 0.85% a day becomes 0.85% × 15.87 ≈ 13.5% a year. That 13.5% is the annualized volatility, and it is the number that would sit next to India VIX for comparison.

The reason you use the square root and not the number itself trips up almost everyone at first. If you hold a position for two days instead of one, your expected dispersion does not double — it grows by √2, about 1.41 times. Four days, not four times but twice. The market's movement partly cancels itself day to day rather than piling up in a straight line, so uncertainty accumulates with the square root of time. Multiplying a daily volatility by 252 would give a wildly overstated 214% a year; multiplying by √252 gives the correct 13.5%.

Not to be confused with: The scaling of returns with the scaling of volatility. Expected return grows linearly with time — hold twice as long, expect twice the drift — but volatility grows with the square root of time. People who annualise a volatility by multiplying by the number of periods, the way you would annualise a return, overstate it enormously. Variance is what scales linearly; volatility is its square root.

Why doubling the horizon does not double the volatility

Volatility grows with the square root of time, not with time

Expected dispersion of NIFTY plotted against holding period, alongside the straight line you would get if it scaled linearly.

The dispersion curve bends below the straight line at every horizon beyond the first, and the gap widens the further out you go. The picture proves that time does not add risk in equal instalments: the second day contributes less new uncertainty than the first, because a √t curve is always flatter than the t line it starts on. Anyone who scales volatility linearly is reading off the wrong curve entirely.

Professional explanation

The square-root-of-time rule, and the assumption underneath it

The rule is short — annualised volatility equals period volatility times the square root of the number of periods in a year — but it rests entirely on one assumption: that returns are independent and identically distributed from one period to the next. If daily returns are independent, their variances add, so the variance over n days is n times the daily variance, and the volatility, being the square root, is √n times the daily volatility. Every annualisation you will ever see is this identity applied. It is exact when returns are genuinely independent and merely approximate when they are not — and real market returns are not perfectly independent, which is the crack the rule eventually breaks along.

Doubling the horizon multiplies dispersion by 1.41, not 2

This is the practical face of the rule and the single most useful thing to internalise. Because dispersion scales with √t, extending a holding period from one day to two multiplies the expected range by √2 ≈ 1.41, not by 2. Going from a month to a year — twelve times as long — multiplies it by √12 ≈ 3.46, not by 12. This is why a 13.5% annual volatility corresponds to only about 3.9% over a single month (13.5% ÷ √12) and about 0.85% over a day. The counter-intuitive consequence is that most of a year's uncertainty is not concentrated at the end; risk accumulates fast at first and then slows, because each additional day adds a fixed lump of variance but a shrinking increment of volatility.

252 trading days versus 365 calendar days — the deliberate inconsistency

Here is the detail that makes a reader's spreadsheet disagree with a broker's screen more often than any other. Volatility is annualised on 252 days, not 365, because a closed market does not move: a weekend contributes no variance, so only trading days count. But time-to-expiry inside an option pricing formula is measured on 365 calendar days, because interest accrues on weekends even though prices do not. So the same option involves both conventions at once — 252 for scaling the volatility, 365 for discounting the time — and they are not reconciled because they are answering different questions. One counts days the market can move; the other counts days money earns interest. The inconsistency is standard, intentional, and the usual reason two correct-looking calculations produce two different numbers.

Where the square-root rule quietly breaks

The rule assumes independence, and it fails precisely when returns are autocorrelated — which is exactly when people lean on it hardest. In a trending market, an up day is more likely to be followed by an up day; the moves reinforce rather than cancel, variance grows faster than linearly, and √t understates the true multi-day risk. In a mean-reverting market the opposite holds: moves partly reverse, variance grows slower than linearly, and √t overstates the risk. A trader annualising a calm, range-bound week's volatility to project the year will overstate it; a trader annualising a quietly trending month will understate it. The rule is a convention that assumes the market has no memory, applied to a market that demonstrably does.

Why the convention still earns its place

Given all that, why annualise at all? Because without a common horizon, no two volatilities are comparable. A daily standard deviation, a weekly one and a monthly one are three different numbers describing the same asset, and only by rescaling them all to a year can you see that they agree — or notice that they do not, which is itself information. When a stock's daily-sampled annualised volatility differs materially from its monthly-sampled annualised volatility, the square-root rule has just told you the returns are autocorrelated, because under independence the two would match. The convention is imperfect as a risk estimate and invaluable as a common language, and its failures are diagnostic rather than merely annoying.

Variance is the additive quantity; volatility only looks additive

Cumulative variance and cumulative volatility of NIFTY over a run of trading days.

Variance climbs as a straight line — each independent day adds the same fixed amount — while volatility, its square root, curves and flattens. This is why every honest volatility calculation is done in variance and converted back only at the end. The straight line is the one that adds up; the curve is the one you quote.

Formula

Annualising a period volatility

σ_annual = σ_period × √N

N is the number of periods of that length in one year. For daily volatility annualised on trading days, N = 252 and √252 ≈ 15.87. For weekly, N = 52; for monthly, N = 12. The rule scales variance linearly (σ² × N) and volatility with the square root, and it is exact only when returns across periods are independent.

σ_annualThe annualised volatility — the one-year figure being produced, expressed as a percentage.
σ_periodThe volatility measured over one period: the standard deviation of that period's returns (0.0085 = 0.85% for a NIFTY day).
NThe number of periods in one year. 252 trading days, 52 weeks, or 12 months, depending on the sampling frequency of σ_period.
√NThe square root of N — the scaling factor. √252 ≈ 15.87, √52 ≈ 7.21, √12 ≈ 3.46. This is the factor volatility scales by, not N itself.

The variance identity it comes from

σ²_annual = σ²_period × N

Read this first and the square-root form follows by taking the root of both sides. Variance — the square of volatility — is what scales linearly with the number of independent periods; volatility inherits the square root. Working in variance and converting back at the end is the way to avoid every square-root-of-time mistake.

How to annualise a volatility, step by step

Compute the standard deviation of returns at your sampling frequency — daily log returns of NIFTY, for example, giving a daily volatility such as 0.85%.
Decide N: the number of those periods in a year. For daily data use 252 trading days, not 365, because a closed market contributes no movement.
Take the square root of N. For daily data that is √252 ≈ 15.87 — the factor you multiply by, never N itself.
Multiply: σ_annual = σ_period × √N. Here 0.85% × 15.87 ≈ 13.5% annualised.
Keep your day-count consistent with its purpose. If you then feed this volatility into an option pricing formula, remember the formula measures time-to-expiry on 365 calendar days even though you annualised the volatility on 252 — the two conventions coexist deliberately.
Cross-check by annualising the same asset at a second frequency — say weekly, with √52. If daily-annualised and weekly-annualised volatility disagree materially, the returns are autocorrelated and the square-root rule is only approximate for this asset.
Report the frequency and day-count alongside the number. "13.5%, from daily returns, 252-day annualised" is unambiguous; "13.5%" alone hides the two choices that produced it.

Practical example

NIFTY worked example

NIFTY at 24,000. Over the past several weeks the standard deviation of its daily log returns works out to about 0.85% per day — an ordinary reading for a calm index. To put this on the same scale as India VIX, annualise it: multiply by √252. Since √252 ≈ 15.87, the annualised volatility is 0.85% × 15.87 ≈ 13.49%, call it 13.5%. Interpret the number: it says that if today's daily behaviour persisted unchanged for a year, NIFTY's returns would have a standard deviation of about 13.5% — comfortably in line with India VIX around 13. Now notice what you must not do. Multiplying 0.85% by 252 gives 214%, which is not a volatility of anything; it is the answer to a question nobody asked, and it is the most common annualisation error there is. The square root is not a refinement — it is the whole rule.

BANKNIFTY worked example

BANKNIFTY at 52,000 shows the same arithmetic delivering a different verdict, and exposes the sampling trap. Suppose BANKNIFTY's daily return standard deviation is about 1.10% — a busier index than NIFTY. Annualised, that is 1.10% × √252 ≈ 1.10% × 15.87 ≈ 17.5%, meaningfully above NIFTY's 13.5%, which is what you would expect from a narrower, bank-heavy index. But suppose you instead sampled weekly and found a weekly volatility of 2.0%; annualised on √52 that is 2.0% × 7.21 ≈ 14.4%, noticeably lower than the daily-sampled 17.5%. The two disagree, and the disagreement is the lesson: BANKNIFTY's daily moves partly reverse within the week, so higher-frequency sampling captures noise that washes out over five days. Which number is right depends on the horizon you care about — and the gap between them is a direct readout of mean reversion that no single annualised figure would have revealed.

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. The square-root-of-time rule is embedded in value-at-risk and margin systems that scale a one-day risk estimate up to a ten-day horizon by multiplying by √10. That scaling assumes independent daily returns, and it systematically understates risk in a trending, autocorrelated crisis — exactly the environment in which the margin is being tested. A ten-day risk built by scaling a one-day number is a convention, not a measurement, and it is least reliable when it is most consequential.

Advantages & limitations

What it is good for

It puts every volatility on one comparable scale. A daily, weekly and monthly standard deviation describe the same asset in incompatible units; annualising them all to a year makes them directly comparable and comparable in turn to India VIX and implied volatility.
It is arithmetically trivial and universally understood. A single multiplication by √N turns any period volatility into the market-standard annual figure, with no model, no fitting and no ambiguity about the method.
It exposes autocorrelation as a diagnostic. Because the rule is exact only under independence, a mismatch between volatilities annualised at two frequencies is a direct signal that returns are trending or mean-reverting — useful information the rule hands you for free.
It scales cleanly in both directions. The same identity that turns a daily volatility into an annual one turns an annual implied volatility into an expected daily or monthly move, which is how traders read an expected range off a quoted number.
It rests on a transparent, checkable assumption. The rule states its own condition — independent returns — so you always know exactly when it applies and exactly what to look for when it does not.

Where it breaks down

It assumes returns are independent across periods, and they are not. Any autocorrelation — trending or mean-reverting — makes variance scale faster or slower than linearly, so the annualised figure is biased in a direction the rule cannot tell you without a second sample.
It understates risk in a trending or crisis market. When up days follow up days, moves reinforce rather than cancel, multi-day variance grows faster than n times the daily variance, and √t scaling gives a number that is too low precisely when accuracy matters most.
It overstates risk in a mean-reverting, range-bound market. When moves partly reverse, √t scaling gives a number that is too high, so annualising a quiet week's volatility projects a year that is more turbulent than the asset's own behaviour implies.
It depends on the sampling frequency you chose. Daily-sampled and monthly-sampled annualised volatilities can differ materially for the same asset, and there is no single correct choice — only a choice matched to the horizon you actually care about.
It clashes with the day-count used elsewhere in the same calculation. Volatility annualises on 252 trading days while option pricing measures time on 365 calendar days, so a figure that is internally correct can still disagree with a broker's number that combined the conventions differently.

Common mistakes

Multiplying by the number of periods instead of its square root. Turning 0.85% daily into an annual figure by multiplying by 252 gives 214%, which is nonsense; the rule uses √252 ≈ 15.87, and forgetting the square root is the most common annualisation error there is.
Annualising volatility the way you annualise a return. Returns scale linearly with time and volatility scales with its square root, because variance is the additive quantity. Applying the linear rule to volatility overstates it enormously.
Mixing 252 and 365 without noticing. Annualising volatility on 365 calendar days, or feeding a 252-annualised volatility into a formula expecting the 365 convention, produces a discrepancy that gets blamed on the model rather than on the day-count.
Trusting a √t-scaled number in a trending market. Scaling a one-day volatility up to a ten-day horizon assumes independence; in a persistent trend the true ten-day risk is higher, and the scaled figure lulls a position into looking safer than it is.
Comparing volatilities sampled at different frequencies as if they were the same measurement. A daily-sampled 17.5% and a weekly-sampled 14.4% for the same asset are not a contradiction to be averaged away — they are evidence of mean reversion, and treating either as the single truth discards that information.
Annualising a handful of returns and quoting the result to three decimals. A standard deviation from a short sample is itself noisy, and √N scaling multiplies that noise by up to 15.87, so a precise-looking annualised figure can rest on almost no real data.

Professional usage

Every volatility a professional quotes is annualised, because annualisation is the lingua franca that lets an equity desk, a rates desk and a risk manager talk about dispersion in one currency. A trader converting a quoted annual implied volatility into an expected daily or weekly move — to size a position or set a stop — is running the square-root rule in reverse: a 13.5% annual volatility implies a daily move of about 0.85% and a monthly move of about 3.9%. Risk teams scale a one-day value-at-risk to regulatory ten-day horizons with √10, fully aware that the scaling assumes independence and understates a trending crisis, which is why the assumption is stress-tested rather than trusted.

Quant researchers use the mismatch the rule exposes as a measurement tool. Because √t scaling is exact only under independence, comparing an asset's volatility estimated from daily returns with its volatility estimated from weekly or monthly returns — a variance-ratio test — directly measures the autocorrelation of returns: a variance ratio above one signals trending, below one signals mean reversion. The square-root-of-time rule, treated as a null hypothesis rather than a fact, becomes a lens on market structure, and its deviations are exactly what the researcher is hunting for.

Key takeaways

Annualized volatility rescales a period volatility to one year by multiplying by √N, where N is the number of periods in a year — √252 for daily data, roughly 15.87.
Variance scales linearly with time and volatility with its square root, so doubling the horizon multiplies dispersion by 1.41, not 2. Multiplying a daily volatility by 252 rather than √252 is the classic error.
Volatility uses 252 trading days because a closed market does not move; option pricing uses 365 calendar days because interest accrues on weekends. The inconsistency is deliberate and the usual reason two numbers disagree.
The square-root rule assumes independent returns and breaks when they are autocorrelated: it understates risk in trending markets and overstates it in mean-reverting ones — precisely when people rely on it most.
A mismatch between volatilities annualised at two frequencies is not an error but a diagnostic: it directly signals that returns trend or mean-revert.

Annualized volatility is the humblest number in the whole subject and the one most quietly full of assumptions. It is a single multiplication, but it hides a theory of how risk accumulates — √t, not t — and two calendars that never reconcile because they answer different questions. Get the square root right and you avoid the 214% embarrassment; understand why the rule breaks and you can read a market's memory off the gap between its daily and monthly volatility. The number is a convention, not a fact, and knowing which is the difference between quoting it and understanding it.

Frequently asked questions

What is annualized volatility?

Annualized volatility is a short-period volatility rescaled to a one-year horizon by multiplying by the square root of the number of periods in a year. It puts daily, weekly and monthly volatilities on one comparable scale, which is why a 0.85% NIFTY day becomes about 13.5% a year.

Why do you multiply by the square root of time?

Because variance, not volatility, scales linearly with time under independent returns. The variance over n days is n times the daily variance, so the volatility — its square root — is √n times the daily volatility. Multiplying volatility by n instead of √n overstates it enormously.

Why is 252 used instead of 365 for annualizing volatility?

Because a closed market does not move, so only trading days contribute variance. There are about 252 trading days in a year once weekends and holidays are removed. Calendar days are used elsewhere — in option pricing time-to-expiry — because interest accrues even when the market is shut.

How do I annualize daily volatility?

Multiply the daily volatility by √252, which is about 15.87. A daily standard deviation of 0.85% becomes 0.85% × 15.87 ≈ 13.5% annualised. Use 252 rather than 365 because volatility only accumulates on days the market is open.

What is the square-root-of-time rule?

It is the identity σ_annual = σ_period × √N, where N is the number of periods in a year. It follows from variance adding linearly across independent periods, so volatility scales with the square root of the number of periods rather than with the number itself.

Does doubling the holding period double the volatility?

No. Because dispersion scales with the square root of time, doubling the horizon multiplies expected dispersion by √2 ≈ 1.41, not by 2. Going from a month to a year multiplies it by √12 ≈ 3.46, not by 12. Risk accumulates fast at first and then slows.

When does the square-root-of-time rule break down?

When returns are autocorrelated. In a trending market, moves reinforce and variance grows faster than linearly, so √t understates risk. In a mean-reverting market, moves partly cancel and √t overstates it. The rule is exact only when successive returns are independent.

Why does my annualized volatility differ from my broker's?

Most often because of the day-count. If you annualised on 365 days and the broker used 252, or you sampled at a different frequency, or one calculation mixed the 252 convention for volatility with the 365 convention for time-to-expiry, the two correct-looking numbers will differ.

Should I annualize on 252 or 365 days?

Use 252 trading days to annualise a volatility, because the market only moves on trading days. Reserve 365 calendar days for the time-to-expiry input inside an option pricing formula, where interest accrues over weekends. The two conventions coexist on purpose and answer different questions.

What is a typical annualized volatility for NIFTY?

NIFTY's annualized volatility usually sits somewhere around 11% to 16% in ordinary conditions, broadly matching India VIX near 13. A daily standard deviation of about 0.85% annualises to roughly 13.5%. Genuine stress periods push the figure well above that range.

Can I annualize weekly or monthly volatility?

Yes. Multiply a weekly volatility by √52 ≈ 7.21 and a monthly volatility by √12 ≈ 3.46. If the daily-annualised and weekly-annualised figures for the same asset disagree, that is a signal the returns are autocorrelated rather than independent.

Why is variance additive but volatility is not?

Because variance is the average squared deviation, and for independent periods the squared deviations add up. Volatility is the square root of variance, and square roots do not add. This is why every careful calculation works in variance and converts back to volatility only at the very end.

What goes wrong if I multiply daily volatility by 252?

You get a number roughly 15.87 times too large — 0.85% × 252 = 214%, which is not the volatility of anything. Annualising uses the square root of the period count, not the count itself, and skipping the square root is the single most common annualisation mistake.

Does annualized volatility predict the annual range?

It describes a one-standard-deviation band under the assumption that today's behaviour persists and returns are independent. A 13.5% figure means about a two-in-three chance the year's return lands within ±13.5%, if nothing changes — which, over a full year, a great deal usually does.

How does annualized volatility relate to the expected move?

Run the rule in reverse. Divide an annual volatility by √N to get the move over a shorter horizon: 13.5% ÷ √252 ≈ 0.85% for a day, 13.5% ÷ √12 ≈ 3.9% for a month. This is how traders read an expected range off a quoted annual number.

Is annualized volatility the same as implied volatility?

Not necessarily. Implied volatility is already quoted on an annualised basis, but it is a price backed out of options, not a historical measurement. Annualized volatility usually refers to a realised, backward-looking figure computed from returns and then scaled to a year. Both live on the same annual scale.

Why does higher-frequency sampling sometimes give a higher volatility?

Because high-frequency returns capture short-term noise and reversals that partly wash out over longer windows. If daily-sampled annualised volatility exceeds monthly-sampled, the asset's moves are mean-reverting; the gap between the two is a direct measure of that reversal.

How many trading days are in an Indian market year?

About 252, once weekends and NSE holidays are removed from the 365 calendar days. The exact count varies by a few days each year, but 252 is the universal convention for annualising volatility, and √252 ≈ 15.87 is the scaling factor that follows from it.

Does the rule work for value-at-risk scaling?

It is widely used to scale a one-day VaR to a ten-day horizon by multiplying by √10, but only approximately. The scaling assumes independent daily returns, so it understates risk in a trending, autocorrelated crisis — which is exactly the scenario the ten-day figure exists to capture.

Can annualized volatility be compared across different assets?

Yes, that is much of its purpose, but with care. Two assets annualised on the same convention and frequency are directly comparable, but a raw comparison ignores that each may have different autocorrelation. Comparing NIFTY at 13.5% with BANKNIFTY at 17.5% is meaningful; both are annualised the same way.

Should I trust an annualized figure from only a few days of data?

Be cautious. A standard deviation from a small sample is noisy, and √252 scaling multiplies that noise by nearly 16, so a precise-looking annualised number can rest on very little. More sampling stabilises the estimate but also mixes in older regimes that may no longer apply.

Voice search & related questions

Natural-language questions people ask about annualized volatility.

How do I turn a daily volatility into a yearly one?

Multiply it by the square root of 252, which is about 15.87. So a 0.85% daily standard deviation for NIFTY becomes roughly 13.5% a year. The key is the square root — you use √252, not 252, because volatility grows with the square root of time.

Why not just multiply by the number of days?

Because that answers the wrong question. Multiplying volatility by 252 gives 214%, which is meaningless. Only variance scales linearly with time; volatility is its square root, so it scales with √252. Doubling the horizon multiplies dispersion by 1.41, not 2.

Why are there two different day counts floating around?

Because they measure different things. Volatility annualises on 252 trading days, since a shut market cannot move. But option pricing measures time-to-expiry on 365 calendar days, because interest keeps accruing over the weekend. Both are correct at once, and that is why a spreadsheet and a broker screen can disagree.

When can I not trust the square-root rule?

When the market has memory. If it is trending, moves reinforce each other and the rule understates the real multi-day risk. If it is mean-reverting, moves partly cancel and the rule overstates it. The √t rule assumes each day is independent, which real markets only approximately are.

My daily and monthly volatility numbers don't match — did I mess up?

Probably not. If daily-sampled and monthly-sampled annualised volatilities differ for the same asset, that is the square-root rule telling you the returns are autocorrelated. A higher daily-sampled figure points to mean reversion; a higher monthly-sampled one points to trending. The gap is information, not a bug.

How do I get from an annual volatility back to a daily move?

Divide by the square root of the number of periods. A 13.5% annual volatility divided by √252 gives about 0.85% for a day, and divided by √12 gives about 3.9% for a month. Running the rule in reverse is how you read an expected range off a quoted annual number.

Is a higher annualized volatility always worse?

It just means more expected dispersion, which is good or bad depending on your position — welcome if you are long options, unwelcome if you sold them. And a high number computed from a short, noisy sample may not be reliable at all, since √252 scaling magnifies whatever noise was in the estimate.

Sources & references

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