Garman-Klass Estimator

Garman-Klass Estimator measures volatility from the whole bar rather than closes alone: each bar contributes half the squared log high/low range minus (2 ln 2 − 1) times the squared open-to-close move, and the window average is annualized into the line. It is a faithful build of the Garman-Klass estimator, with a close-to-close line on identical settings so the blind spot and the efficiency edge are both visible.

How to Trade the Garman-Klass Estimator?

• Compression: the percentile rank crossing under the low threshold flags a quiet stretch — a rank-based squeeze screen for lulls that often resolve into wider movement.
• Expansion: the rank crossing the high threshold confirms volatility is broadening.
• Gap Blind Spot: Garman-Klass far below close-to-close says the action is happening between sessions, in gaps it never samples — trust the pair, not the level.
• Session Churn: the estimate far above close-to-close flags heavy intrabar trade that the closing prints barely register.

Garman-Klass Estimator Settings

• Window Length (default 20): bars the per-bar variance is averaged over; the estimator's efficiency makes 10–30-bar windows usable where close-to-close is still noisy.
• Annualize (default enabled) with Bars Per Year (default 252): scales readings to annualized percent — match the constant when reconciling with other tools.
• Close-to-Close Comparison (default enabled): the same-window benchmark line.
• Divergence Alert Level % (default 25): how far the two lines must diverge before the diagnostic alerts fire.
• Percentile Lookback (default 100) with High Percentile (80) and Low Percentile (20); Show Percentile Levels (default disabled) plots the thresholds as dotted levels.
• Show Dashboard (default enabled); line colors and Gradient Fill are cosmetic.

Frequently Asked Questions

How does it compare with the Parkinson estimator?

The Parkinson Estimator uses the high-low range alone; Garman-Klass adds the open-to-close leg with the paper's optimal weights, extracting more information from the same bar. Both share one limitation: neither term sees the close-to-open gap.

Why is my reading lower than close-to-close volatility?

When Garman-Klass sits far under the close-to-close line, the variance is landing in the overnight span the formula omits. The dashboard's GK / Close-to-Close ratio quantifies it.

Does it work intraday?

Yes — the formula is bar-agnostic — but set Bars Per Year to bars per session times trading days, or annualized levels will be wrong. The percentile thresholds keep working regardless — they rank the reading against its own history.