Unfortunately, I no longer have these data. I locked them away in my university account years ago, and haven't looked at them for a long while. They've probably been purged from the account already I'm guessing. However, I do recall that the distribution isn't all that "normal." IF I remember correctly, the Chi-Squared Test of fitness only scored around 0.85. I believe Peters (Fractal Market Analysis) came up with some Fractal Distribution. In all seriousness, I think this is just another Black Scholes at best, and Wizetrade behind equations at worst.
JS11374, SIGH! I hear ya! I've done more than my fair share of simulations, gotten the conclusions, lost the original data/files, and regretted it later. I do suspect you are right about the usefulness of the Hurst exponent. The fact that the calculation uses Max-Min would make the Hurst exponent too sensitive to random patterns in any extreme outliers IMO. And if you attempt to gather enough data to "smooth" over the outliers, you end up with too large a lag in the result. Who wants to gather 10 years of 1-minute data ( a million samples) to successfully conclude that the markets were trending 5 years ago. Its the curse of living with platykurtosis on the right edge of the chart. -Traden4Alpha
A relatively inexpensive program for calculating Hurst exponent as well as a whole series of other useful stats is: Chaos Data Analyzer Pro - Physics Academy Software
JS When I performed these tests about 5 years ago, I used DAILY Tick Data. The data went as far back as 1969 on 37 futures markets. I used a time frame of one day up thru 6 months. If you are a very short term trader then I assume this would not be too helpful for you. JJ
Actually, I don't think the tick data would be any value. It then becomes a matter of both under and over sampling (sound strange, doesn't it). Under in the sense that you are still only getting a local subset of the global price set (a set that if you actually have, you wouldn't need to do anything else because you would know what tomorrow's price is). It's over sampling in the sense that tick data introduces a whole lot of unnecessary noise into analysis. Other than some fundamental statistic analysis on transaction data (interesting, but useless because of slippage and commission), most academic studies stay away from tick data because by you cannot surmise any global information from it. Trying to obtain short term information from tick data using that much data simply introduces yet another whole set of statistical issues. Some may argue that I'm just being too stringent on this whole matter, but thing is this is what differentiates statistical studies from technical analysis.
JS You may be confused by my saying "Tick Data". That was the vendor source of my daily closing prices. JJ