Combining Hurst Estimators to Measure NQ Intraday Memory

Summary

This article presents a method for estimating long-range dependence in NQ one-minute futures returns and using it to characterize market regimes. It combines three estimators: rescaled range, aggregated variance, and absolute moments. Each estimator fits a log-log relationship across scales and receives a confidence weight based on the number of usable regression points. Their weighted blend is intended to be more informative than relying on one estimate, since the methods have different sensitivities to short-range dependence, trends, structural breaks, and heavy tails.

An empirical study of Globex data reports that estimates generally sit near the random-walk boundary, with the pooled blend slightly above it and rolling averages below it. The author cautions that the blended measure did not significantly predict intraday trending in the sample. It is better treated as a description of position relative to the random-walk boundary than as a standalone directional signal. Estimates require sufficient data, and pre-open and post-open observations should not be mixed.

Key ideas

• Persistent Hurst estimates above 0.5 are associated with long-term dependence, while estimates below 0.5 indicate anti-persistence.
• Rescaled range, aggregated variance, and absolute moments have different assumptions and failure modes.
• The proposed composite weights each estimator according to the availability of valid regression scales.
• The NQ study found estimates near the random-walk boundary and did not establish significant prediction of intraday trends.
• Returns, adequate sample length, and consistent session windows are important inputs to the method.

Tags

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