Local Fractal Index Estimation Compared with the Hurst Exponent

Summary

The article compares the Hurst exponent with a local fractal index for characterizing financial price series. It describes the Hurst-based relation between persistence and fractal dimension, but notes that reliable estimates generally need long samples and can be affected by distributional assumptions and changing market behavior. It then explains a minimum-cover method: divide a series into intervals, measure the summed price variation at different scales, fit a line to the log-scale relationship, and use its slope to estimate the index. The method is presented as suitable for local calculations on shorter windows.

Examples include historical Lukoil prices and a rolling-window analysis of Alcoa data, with reported fits and estimates used to illustrate differences between the measures. The article proposes interpreting index values below, above, or near a midpoint as trend-like, flat or anti-persistent, and random behavior, respectively. These classifications are heuristic: the examples do not establish forecasting profitability, estimates depend on window and scale choices, and local properties can change over time. The article’s own examples show that the Hurst-based and cover-based estimates can disagree.

Key ideas

• The Hurst exponent can require long samples, limiting its usefulness for rapidly changing local conditions.
• The minimum-cover approach estimates a fractal index from price variation measured across multiple scales.
• A log-scale regression slope supplies the index estimate, and the article illustrates the method on historical equities.
• Fractal classifications describe statistical structure but do not by themselves demonstrate profitable forecasts.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.