Volatility Signature Plot

Volatility Signature Plot draws the econometric sampling diagnostic in the space beside price, a first for the library. Realized variance is measured from squared intraday returns at up to six intervals and averaged over completed days, plotted as a curve against interval — the volatility signature plot. A frictionless market would draw a flat line, so the bend at fine intervals is the diagnosis — colored point by point and summarized on a dashboard with a suggested sampling interval.

How to Trade the Volatility Signature Plot?

• Upward fine-end slope: fine-sampled variance runs above the coarse level — bid-ask bounce inflating high-frequency returns; alerted at onset.
• Downward slope: fine-sampled variance runs low — stale quotes or thin trading smoothing the feed; alerted.
• Suggested interval: the finest point still within tolerance of the coarse level is highlighted — the sampling that keeps the most data without absorbing microstructure bias.
• Flat signature: the whole grid inside tolerance — realized volatility insensitive to the sampling choice; alerted when it takes hold.

Volatility Signature Plot Settings

• Interval 1 (min)–Interval 6 (min) (defaults 1, 3, 5, 15, 30, 60), each with an enable toggle: the sampling grid.
• Days To Average (default 20): completed days per point; single-day plots are noise, and the dashboard reports the count used.
• Exclude Overnight Return (default on): keeps overnight jumps out of the intraday question.
• Flatness Tolerance % (default 10): drives both the bias diagnosis and the suggested interval.
• Output Units (default Annualized Volatility %) with Days Per Year (default 252).
• Display: Point Spacing (bars) (default 8), Rightward Offset (bars) (default 5), Show Flat Reference Line, Show Point Values, Show Dashboard (all on, Top Right, Small).

Frequently Asked Questions

What do I do with the suggested interval?

Make it the operating resolution for realized-volatility work: tools like the Volatility Estimators indicator inherit whatever bias lives in their input returns, so sampling near the finest flat point keeps readings clean.

Why exclude the overnight return?

Because the plot asks an intraday sampling question and the close-to-open jump belongs to a different volatility component. Gap detection is automatic — continuous sessions keep their day-boundary return.

Why do fine intervals show fewer days?

Available intraday history is shortest at the finest granularities, so those points may average fewer days than the target — the dashboard flags short counts. Such values are legitimate, just noisier.