Concept

Volatility Signature Plot

Volatility Signature Plot is a Volatility concept.

What is a volatility signature plot?

A volatility signature plot graphs realized volatility against the sampling frequency used to compute it. You measure realized variance from squared returns sampled every 1 minute, 5 minutes, 15 minutes, and so on, average each estimate over many days, and plot the result as a function of the sampling interval. In a frictionless market the line would be flat, since true variance does not depend on how often you look. Real markets bend the line at fine intervals, and the shape of that bend is the signature.

The tool comes from the realized-volatility literature in econometrics around 2000, associated with the work of Torben Andersen, Tim Bollerslev, Francis Diebold, and Paul Labys. Its purpose is diagnostic: microstructure frictions such as bid-ask bounce induce negative autocorrelation in very high-frequency returns, which inflates measured variance as sampling gets finer. Illiquid instruments with stale or sluggish quotes can show the opposite, a downward bias at fine intervals.

Practitioners care because realized volatility is only as good as its sampling choice. The signature plot makes the bias visible and points to the coarsest region where the line flattens, historically often in the vicinity of 5-minute sampling for liquid instruments, as a sensible operating point. It is the standard sanity check before feeding realized measures into forecasting models, execution analytics, or any of the volatility estimator comparisons that assume clean inputs.

How it's calculated

Realized variance computed at a grid of sampling intervals, averaged across days and plotted against the interval.

r_(i,h) = ln(P_(i*h) / P_((i-1)*h))
RV_d(h) = sum(r_(i,h)^2) over the M(h) intraday returns of day d
SIG(h) = (1/D) * sum(RV_d(h)) over d = 1 to D
plot SIG(h), or its annualized square root, against h
h: sampling interval (e.g. 1, 5, 15, 30 minutes)
P_(i*h): price at the i-th grid point of the day at interval h
r_(i,h): intraday log return at interval h
M(h): number of returns per day at interval h
RV_d(h): realized variance of day d at interval h
D: number of days averaged
SIG(h): the signature value at interval h

Averaging over many days (often 20 or more) is essential; a single day's plot is dominated by noise.

Prices can be sampled from trades, midquotes, or previous-tick interpolation, and the choice changes the fine-interval behavior.

Overnight gaps are usually excluded and handled separately, since they are not part of the intraday sampling question.

How traders use it

  • Choosing a sampling frequency: the standard read is to pick the finest interval at which the plot has flattened, keeping as much data as possible without absorbing microstructure bias into the volatility estimate.
  • Diagnosing data quality: an upward-sloping signature at fine intervals on trade data flags heavy bid-ask bounce, while a downward slope often betrays stale quotes or thin trading, useful before trusting any high-frequency analytics on that instrument.
  • Comparing instruments and venues: signature plots for the same asset from different data feeds or exchanges reveal which source is cleaner at a given resolution.
  • Its limitations: the flat region can be ambiguous or absent for illiquid names, the plot is an average that hides day-to-day variation, and it diagnoses bias without removing it; noise-robust estimators exist beyond the scope of the plot itself.

Signature plot vs related volatility views

Realized Volatility: Realized volatility is the quantity being estimated; the signature plot is a diagnostic of how that estimate changes with sampling choice. One is the measurement, the other the calibration chart for making it.

Volatility Term Structure: The term structure varies the forecast or maturity horizon while holding the measurement fixed. The signature plot varies the sampling interval of the measurement itself while the period measured stays the same. The axes look similar; the questions are unrelated.

Volatility Cones: Cones summarize the historical distribution of volatility across window lengths for valuation context. The signature plot is not about history or valuation at all; it is about measurement bias at fine time scales.

Concept family

Volatility

57 concepts mapped · 57 in the Library

Volatility Signature Plot FAQ

Why does measured volatility rise at very fine sampling intervals?

Mostly bid-ask bounce: trades alternate between bid and ask, creating back-and-forth price changes that are noise, not true variance. Squaring and summing those tiny reversals inflates realized variance as the interval shrinks.

Is 5-minute sampling always the right choice?

No. It became a common convention because signature plots for liquid instruments often flatten around there, but the right interval is whatever your instrument's own plot supports. Thinner names may need 15 or 30 minutes; very liquid futures can sometimes go finer.

Can the signature plot slope downward instead of up?

Yes. Stale quotes and infrequent trading smooth prices at fine intervals, inducing positive autocorrelation that biases variance downward. A downward-sloping signature is a common fingerprint of illiquidity.

Do everyday chart traders need signature plots?

Rarely for discretionary charting, since bar-based tools like ATR sidestep the issue at coarse resolutions. They matter as soon as you compute realized volatility from intraday returns for sizing, forecasting, or research, where sampling bias silently distorts results.

Build Volatility Signature Plot your way.

Quant writes, tests, and refines it with you — then it runs on LuxAlgo charting or ports to TradingView.