Market regime (trending vs mean-reverting)
A market regime is a period during which a price series behaves in a consistent way along some measured dimension, most often persistence (trending versus mean-reverting) or volatility (agitated versus calm). Regime measures classify the recent window so that a strategy can be applied only in the conditions it was built for.
Senzoukria · Glossary · Updated September 2026
Dimensions of a regime
Each axis is independent. A market can be trending and calm, or mean-reverting and agitated. Strategies usually depend on one or two axes: a fade needs reversion, a breakout rule needs persistence and often an expansion in volatility.
| Axis | States | Typical measure |
|---|---|---|
| Persistence | Trending, random, mean-reverting | Hurst exponent, autocorrelation of returns |
| Volatility | Agitated, ordinary, calm | Ratio of short to long ATR, realised volatility |
| Options positioning | Dampened, amplified, on the flip | Dealer gamma exposure versus spot |
| Latent state | Model-defined, e.g. crisis vs normal | Hidden Markov model on returns |
Measuring a regime honestly
- Every regime measure is computed over a window; the classification describes that window and lags any change.
- Ratios are comparable across instruments where raw values are not: a short-to-long volatility ratio near 1 means the same on any contract.
- When the denominator is zero or the window is incomplete, the measure is undefined; it should display no value, not a neutral one.
- A regime tag is a filter for when to apply a rule, not a signal in itself.
In Senzoukria
The "Volatility Regime" indicator (catalog id volatility-regime) divides a fast ATR by a slow ATR, defaults 14 and 100, and reads above 1 as an agitated regime, below 1 as calm and near 1 as ordinary. Being a unitless ratio it can be compared across instruments and periods. When the slow ATR is zero the indicator returns no value, since 0 ÷ 0 is not a calm regime but an absence of information.
For persistence, the "Hurst Exponent (R/S)" indicator and the Hurst MR Signal's status badge show a regime tag against the 0.5 threshold. The GEX workspace labels the options regime as "Dampened", "Amplified" or "On the edge" relative to the gamma flip, with the caveat that this reading assumes dealers are net short options; it depends on an options data source being configured. The scripting template "Mean reversion under regime (HMM + Hurst)" combines a hidden Markov state with the Hurst filter, in Python.
Common mistakes
- Assuming one regime tag covers every timeframe; a 5-minute chart and a daily chart can disagree.
- Trading the regime change itself as if the measure could detect it on the bar it happened.
- Using raw ATR to compare agitation across instruments, where only a ratio is comparable.
- Reading a missing regime value as 'ordinary' rather than as 'no information'.
Related
- Hurst exponent and mean reversion guide
- Volatility Regime indicator
- Hurst exponent
- Mean reversion
- Zero gamma flip explained
This page in other languages
Frequently asked questions
- Why use a ratio of two ATRs instead of ATR alone?
- ATR is in price units and depends on the instrument and the era, so a value of 12 means nothing without context. A fast ATR over a slow ATR is unitless and compares the moment with its own normal. The slow window has to be long enough not to follow the fast one, otherwise the ratio collapses toward 1.
- Can a regime filter be backtested?
- Yes, as part of the strategy: the filter's window and thresholds are parameters, and they must be chosen on training data and evaluated out of sample like any other. A filter tuned on the same data it is judged on will look better than it is, which the walk-forward tools in the backtest panel are designed to expose.
- Is the GEX regime the same kind of measure?
- It classifies a different axis. Persistence and volatility regimes are computed from price alone; the GEX regime is derived from the options chain and dealer positioning, and it is conditional on an options data source. The three can be read together but are not substitutes.