Calculations for the Hurst exponent

Has anyone averaged out the total number of bars available to find the hurst exponents overall performance? Then rank each pair to it's highest exponent closest to +1?
I think if some could get the math correct and the total average it would make a great contribution to the indicators available on FF.

Been considering this myself after reading Ernest Chan - Algorithmic Trading.

There is a MATLAB library referenced in the book for finding the "generalised Hurst exponent of a stochastic variable". The source code can be found here http://uk.mathworks.com/matlabcentra...ent/genhurst.m

I'm a coder, but I currently have zero experience of MATLAB code. I doubt it would be too difficult to convert into MT4.

The primary use of this I would think would be to use in an algorithm to determine whether to favour mean reversion or momentum strategies. The result would of course be lagging, and highly subjective to the lookback period used. There would be little point performing parameter optimisation on a particular dataset for the best lookback period, since this would then introduce data-snooping bias. Some fixed value would have to be determined, and then the strategy optimised in a different manner.

Talking to myself here I know, but interested in this idea - going to fire up Matlab over the weekend, stick a debugger on it, and walk through genhurst so I can replicate the code precisely in MT4, and come up with an indi to show the Hurst exponent over time, and then visually inspect it with various lookback periods to see if it is any kind of useful indicator of a price series being stationary or trending at a given point in time.