Concept

Realized Volatility

Realized Volatility, also known as realized variance, is a Volatility concept. A reference entry: the Library explains it rather than implements it.

intraday-summed

What is Realized Volatility?

Realized volatility is the volatility an asset actually printed over a past window, computed from observed returns rather than from option prices. In its strict form, realized variance sums squared returns sampled at a higher frequency than the evaluation window (five-minute returns summed across a day, for example), and realized volatility is the square root of that sum, usually annualized. Because it aggregates squared returns directly, the estimate converges on the true variation of the price path as sampling gets finer, until bid-ask bounce and other microstructure noise begin to dominate.

The strict usage comes from the high-frequency econometrics of the late 1990s, where researchers led by Andersen and Bollerslev showed that summing finely sampled squared returns gives a far sharper daily volatility estimate than any single close-to-close number, with five-minute sampling emerging as the practical convention that balances information against microstructure noise. That literature is why 'realized' means measured-from-the-path rather than merely historical.

On charts the term is used more loosely: most implementations take the standard deviation of daily log returns and annualize it by the square root of periods per year, which is properly close-to-close historical volatility. Both usages are common, so check the formula before comparing numbers. Either way, realized volatility is the backward-looking anchor of the volatility toolkit: it is what sizing models scale against, what regime filters classify, and the measured leg of the spread against implied volatility that option traders watch.

Two behavioral facts drive most of its practical use. Volatility is persistent, arriving in clusters, so the recent reading is genuinely informative about the near future even though it forecasts nothing about direction. And volatility is comparable where price is not: expressing moves, stops, and patterns in units of current realized volatility puts a quiet blue chip and a wild crypto pair on one scale, which is what volatility percentile framing and volatility-targeted risk budgets exploit.

How to read a realized volatility pane

The series plots as a smooth line or histogram under price; reading it is mostly reading its own definition.

  1. 1Check the formula first: summed intraday squared returns versus the standard deviation of daily closes give different numbers with the same name.
  2. 2Check the window and annualization: a 10-day and a 90-day series tell different stories, and annualization conventions (252 versus 365 days) shift levels between markets.
  3. 3Read the level against the instrument's own history, a percentile view, rather than against another asset's numbers.
  4. 4Respect persistence: elevated readings tend to follow elevated readings, so a spike is a regime statement, not a one-bar event.
  5. 5Cross-check against ATR: the range-based sibling includes gaps and intrabar travel that close-based calculations miss, and disagreement between the two is itself informative.
  6. 6For options context, lay the series against implied volatility: the spread between priced and printed volatility is the volatility risk premium.

How it's calculated

The volatility an instrument actually exhibited over a past window, measured from its own returns rather than implied by option prices.

rt=ln⁡(Ct/Ct−1)r_t = \ln\left(C_t / C_{t-1}\right)
RVar⁡=∑t=1nrt2\operatorname{RVar} = \sum_{t=1}^{n} r_t^2
RV⁡=RVar⁡\operatorname{RV} = \sqrt{\operatorname{RVar}}
RV⁡ann=An×∑t=1nrt2\operatorname{RV}_{\text{ann}} = \sqrt{\frac{A}{n} \times \sum_{t=1}^{n} r_t^2}
Sample-stdev variant: RV⁡ann=A×1n−1×∑t=1n(rt−rˉ)2\text{Sample-stdev variant: } \operatorname{RV}_{\text{ann}} = \sqrt{A} \times \sqrt{\frac{1}{n - 1} \times \sum_{t=1}^{n} \left(r_t - \bar{r}\right)^2}
C_t: close of interval t
C_(t-1): close of the previous interval
r_t: log return of interval t
ln(): natural logarithm
n: number of returns in the window (commonly 20 or 30 daily returns)
RVar: realized variance over the window
RV: realized volatility, the square root of RVar
RV_ann: annualized realized volatility
A: return intervals per year (252 for daily data)
r_bar: arithmetic mean of the n returns
t: interval index

High-frequency realized variance sums squared intraday returns (for example 5-minute) within each day; the close-to-close form above is the common charting version.

The zero-mean convention (no r_bar subtraction) is standard for short windows; the sample standard deviation variant subtracts the mean and divides by n - 1.

Multiply by 100 to quote as a percent.

How traders use it

  • Sizing positions: volatility-targeted sizing divides a fixed risk budget by current realized volatility, so exposure shrinks when the tape gets wild and grows when it goes quiet.
  • Classifying regimes: comparing today's reading against its own history (a volatility percentile over the past year, for instance) labels the environment quiet or stormy and tells you which playbook applies.
  • Benchmarking implied volatility: when options price substantially more volatility than the market is realizing, sellers see a premium and buyers a headwind; the spread between the two is itself a monitored quantity.
  • Feeding forecasts: rolling realized measures are the raw input to EWMA and GARCH-style models, which weight recent squared returns more heavily to estimate the next period's volatility.
  • Normalizing everything else: expressing stop distances, breakout sizes, and pattern heights in realized-volatility units makes rules transferable across symbols and eras, the same service ATR provides in range terms.

Realized Volatility vs. neighboring measures

Close-to-close Historical Volatility: Uses one close-to-close return per bar and takes a sample standard deviation over a window. Strict realized volatility instead sums squared intraday returns, extracting far more information from each day; in loose chart usage the two names blur into one.

Implied Volatility: Backward versus forward: realized volatility measures what already happened, while implied volatility is backed out of option prices as the market's priced expectation of what comes next. The gap between them is tracked as the volatility risk premium.

EWMA Volatility: A rolling realized calculation weights every observation in the window equally and drops each one abruptly at the window's edge. EWMA decays weights exponentially, so the estimate reacts faster to fresh shocks and avoids lookback-edge artifacts.

Concept family

Volatility

57 concepts mapped · 57 in the Library

Realized Volatility FAQ

Is realized volatility the same as historical volatility?

In everyday chart usage, yes: both usually mean the annualized standard deviation of past returns. In the stricter academic sense they differ: realized volatility is the square root of summed squared intraday returns, estimating a single day's variation, while historical volatility takes one return per bar over a longer window. Check which formula an implementation uses before comparing values across sources.

How do you annualize realized volatility?

Multiply the per-period figure by the square root of the number of periods in a year: roughly the square root of 252 for daily equity returns, or of 365 for markets that trade every day. The scaling assumes returns are independent across periods, which is an approximation, so treat annualized numbers as a shared yardstick rather than a precise forecast.

Does high realized volatility tell you which way price will go?

No. Realized volatility is direction-blind: it measures the size of moves, not their sign, and strong rallies and sharp selloffs can print similar readings. It earns its keep in sizing, stop width, and regime classification, with directional views coming from trend or structure tools rather than from the volatility number itself.

What sampling frequency does strict realized volatility use?

Five-minute returns became the working convention in the research literature: fine enough to capture most of the day's variation, coarse enough that bid-ask bounce does not dominate the sum. Liquid markets tolerate finer sampling; thin ones need coarser. The trade-off is fundamental, more samples mean more information and more microstructure noise, and refinements like subsampling exist precisely to manage it.

How is realized volatility different from ATR?

Different raw material. Realized volatility is built from returns between prices, squaring and summing them into a statistical estimate in percentage terms. ATR is built from bar ranges including gaps, expressed in price units. ATR sees intrabar travel that close-based measures miss; return-based measures plug directly into statistical machinery. Most desks track both and treat divergence between them as a data point.

Why does volatility arrive in clusters?

Empirically, large moves follow large moves and quiet follows quiet, one of the most robust facts about financial returns. Mechanically, information arrives in bursts, positioning unwinds propagate, and risk models that respond to volatility feed back into trading. For practice the cause matters less than the consequence: current realized volatility is a genuinely useful input for near-term risk, which is exactly why sizing and regime rules key on it.

Turn Realized Volatility into a trading strategy.

Describe your Realized Volatility idea to Quant. It builds the strategy with you and backtests it on real data.