The Hurst Exponent Tells You If a Market Is Trending or Fighting to Come Back
Not all random-looking series are equally random. Some financial series have a hidden directional bias; big moves tend to be followed by more moves in the same direction. Others have a hidden pull toward the mean; when a series gets too high or too low, it tends to drift back toward center. And some are genuinely close to pure random walks, where the past gives you essentially no information about the future.
The Hurst exponent, developed by hydrologist Harold Edwin Hurst in the 1950s while studying the Nile River's flooding patterns, gives you a single number between 0 and 1 that characterizes this behavior. A Hurst exponent near 0.5 means the series behaves like a random walk; no memory, no directional bias. Above 0.5 suggests trending behavior, also called persistence; up moves tend to follow up moves. Below 0.5 suggests mean-reverting behavior, also called anti-persistence; moves tend to reverse.
The FatNarwhal Hurst Exponent tool computes this number for any financial time series and tells you what it implies about the series' behavior.
How to Use It
Go to fatnarwhal.com/hurst and enter a ticker.
The tool measures how fast the series spreads out as you look further apart in time (by default on the log price) and outputs the estimated Hurst exponent along with a regime label: mean-reverting, random walk, or trending. It also plots the log-log scaling chart, which is the underlying calculation visualized; the slope of that line is the Hurst exponent.
Try a few different asset classes and compare. Equity index returns often come in near 0.5 (close to random walk), consistent with efficient market theory. Individual stocks sometimes show mild persistence or anti-persistence depending on the time period and scale. Volatility series and some commodity prices show stronger evidence of long memory. Spread series from cointegrated pairs often show Hurst exponents below 0.5, which is another way of saying they are mean-reverting, which is exactly what makes them useful for pairs trading.
Change the lookback period and see how stable the estimate is. A robust Hurst exponent should not change dramatically when you shift the window.
The Math Behind It
The tool uses the structure-function (lagged differences) method. For each lag τ it takes the root-mean-square of the change in the series over τ steps; for a series with Hurst exponent H, that typical move grows like τ to the power H. H is the slope of a log-log regression:
If H = 0.5, the typical move grows as √τ, exactly as a random walk would. If H > 0.5, it grows faster than √τ, indicating persistence. If H < 0.5, it grows slower, indicating anti-persistence.
The tool labels anything between 0.45 and 0.55 a random walk. It does not report a standard error, so treat small departures from 0.5 as noise unless they hold up across different lookback periods.
When to Use It and When Not To
Use the Hurst exponent as a preliminary diagnostic when deciding whether a trend-following or mean-reversion strategy might work on a particular series. A Hurst above 0.6 is a reasonable signal that trend-following approaches have something to work with historically. A Hurst below 0.4 suggests mean reversion is a more natural fit.
Use it on spread series from pairs trading candidates to confirm that the spread has mean-reverting properties beyond what the cointegration test alone tells you.
The main caution is that Hurst estimates are noisy on short samples; you need at least several hundred data points for a reliable estimate, and even then the confidence interval is wide. And like all historical statistics, the Hurst exponent measures what the series did in the past, not necessarily what it will do going forward. A series that was mildly trending over 2018-2022 might behave differently in the next four years.
Also, different estimation methods (R/S analysis, DFA, variance ratio, structure functions) can produce noticeably different Hurst estimates on the same data. The structure-function method used here is simple and robust, but it is worth knowing that there is no single universally accepted estimator.
Try It
Go to fatnarwhal.com/hurst and run the Hurst exponent on a few different series; a major equity index, a volatile individual stock, a currency pair, and the spread from a cointegration test you ran earlier. Compare the results. The differences in H across series are a picture of how different markets have different underlying dynamics and why no single strategy works everywhere.
One number. A surprising amount of information about how a series moves through time.