Hurst Exponent

Hurst Exponent is the definitive clean build of the rolling Hurst exponent read. Each bar it runs classical rescaled-range analysis on the last 128 increments, fitting log R/S against log subsample size across dyadic scales; the fitted slope is H. Because raw rolling estimates jump, a smoothed series carries the regime read, and a neutral band around the 0.5 benchmark absorbs estimation noise: no regime is called until the smoothed H clears it. The line colors by state, and the dashboard quotes the reading with its method, window and log-log fit R².

How to Trade the Hurst Exponent?

• H above the band: persistent increments, the ground where trend-following approaches have a case.
• H below the band: anti-persistence. Moves lean toward giving part of the ground back, favoring fading and reversion tactics.
• H inside the band: random-walk territory; past direction carries no usable information, so both playbooks stand down.
• Regime alerts: fire on each crossing into trending, mean-reverting or neutral territory.

Like much of the statistics family, the reading is a tendency, not a precise number: a filter over strategies, not an entry.

Hurst Exponent Settings

• Source (default close): series the exponent is estimated on.
• Sample Length (default 128): rolling increments per estimate; longer is steadier but slower to adapt.
• Increments (default Log Returns): or Simple Differences for series that can be zero or negative.
• Smoothing Method (default SMA) with Smoothing Length (default 5): the smoothing that carries the regime read; 1 follows the raw estimate.
• Neutral Band Width (default 0.05): half-width of the noise band around 0.5.
• Dashboard & style: dashboard on, gradient fill on, raw estimate hidden by default.

Frequently Asked Questions

How does the Hurst Exponent differ from Autocorrelation?

Why a neutral band instead of a hard 0.5 line?

Does a high H say which way price will go?