Measuring ATR Breaches and Volatility Clustering
Summary
The indicator measures how often price exceeds a level derived from ATR, then compares a recent breach frequency with its longer history. Its default method tests today’s high and low against bands set using the previous bar’s ATR, so the threshold is fixed before the move. A second method compares the current bar’s true range with ATR, measuring bar size rather than a move beyond a prior expectation. The author also uses a z-score to flag unusual recent breach rates and estimates whether breaches tend to follow one another.
Examples from ten years of daily data across eight instruments are offered as evidence of clustering, with shuffled-series comparisons used to illustrate a baseline. The article emphasizes that these estimates are fragile: short windows may contain few or no breaches, while autocorrelation and non-normality make the z-score a heuristic rather than a significance test. The readings are not directional signals; proposed uses include adjusting position size or stop distance and assessing breakouts. Indicator code is included, but its calculations require enough history to warm up.
Key ideas
- A prior-bar ATR band tests whether price exceeded a level fixed before the current move.
- Recent breach frequency can be compared with its own longer-run distribution to flag unusual conditions.
- Conditional breach probability measures clustering, but short windows can leave it based on very few observations.
- The article treats z-score thresholds as heuristics because breach rates are autocorrelated and non-normal.
- Breach statistics describe volatility behavior, not the direction of the next price move.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.