Lookback period for Hurst Exponent calculation

Hi,



I've been trying to understand the Hurst Exponent calculation on real asset price time series and I noticed that the data span from 2000-2019. The calculation uses all the data in 20 years. 



My question is: Shouldn't we calculate the Hurst Exponent based on the look-back period? The conclusion that SPX is trending on a 20-year timescale isn't really useful for a momentum strategy with a 12-month look-back period and a 1-month holding period. A price series may display long-term trending properties but on shorter timeframes it may display mean-reversion properties. Does this calculation in the Python Notebook take into account these factors?



Thanks in advance!

Hello Yefeng,



Hurst value is to identify the nature of the time series. We are only interested in securities which are trending over a long period and therefore calculated Hurst over a long period, namely 20 years. To make sure that it is not mean-reverting in the short term we did the correlation analysis subsequently - over the look-back range. So if it is mean-reverting in the short term then that lookback and hold period combination will have a negative correlation.