Mean reversion

The idea and its measurement

A mean-reverting series is one where a move away from its mean makes a move back more likely than a further move away. The mean has to be defined: a rolling average of price over N bars is the common choice, and the distance from it is normalised by the rolling standard deviation to give a z-score. A z-score of +2 says price is two standard deviations above its recent mean; in a mean-reverting regime that is an extension to fade, in a trending regime it is a breakout to follow. The number alone does not say which.

• Reference: rolling mean, VWAP, anchored VWAP or a model fair value.
• Extension: the z-score (price − mean) ÷ standard deviation over the window.
• Regime check: a Hurst exponent below 0.5 over the same window indicates the series has been reverting.

Anatomy of a mean reversion trade

The guide on this site describes a typical build in four steps. First, only look for trades when the regime measure indicates reversion. Second, wait for a z-score extension away from the fair-value line, the ±Kσ envelope. Third, fade the extension toward the mean, with the stop beyond the envelope and the target at or before fair value. Fourth, stand aside when the regime measure crosses back into trending territory, because the condition that made the trade valid is gone. The first two steps can be automated by an indicator; the entry trigger, size and exit discipline remain the trader's.

• Stops for a fade sit outside the envelope, where the reversion thesis is falsified.
• Targets sit at or before the mean, since the mean itself moves toward price as the trade runs.
• Time stops matter: a reversion that has not started within the expected horizon is a regime signal.

In Senzoukria

The "Price Z-Score" indicator (catalog id price-zscore) computes the rolling z-score of the close over N bars, defaulting to 20, with a population standard deviation and no session reset; a flat window returns no value rather than zero. The Hurst MR Signal from the mean reversion guide draws a fair-value line with a dashed ±Kσ band, triangles at fade signals, tags where the hypothetical trade came off (tp, sl, to) and a status badge with the live Hurst and z-score against their thresholds.

In the Scripting screen, the template "Mean reversion under regime (HMM + Hurst)" combines a hidden Markov model that rules out the crisis regime, a z-score that says when to enter and toward which price, and the Hurst exponent as confirmation; it is Python only because it needs numpy. None of these place orders on their own.

Common mistakes

• Fading every extension without a regime filter, which fails on the first trend day.
• Choosing the z-score window after seeing the chart and calling the result a test.
• Placing the target at the mean as of entry rather than the moving mean.
• Treating an absent z-score (flat window, warm-up) as zero and reading it as no extension.