[math]

Anyone thinking about doing doing some technical analysis may well become less enthusiastic about the idea if they first played around with generating some random time-series with statistics similar to the real thing and then dwelling on the fact that they look pretty close to the real thing. GARCH(1,1) isn't a bad place to start (although it's not perfect). I just put up some R code on github for this here: https://github.com/mhowlett/garch11 which I was experimenting with in the past. I can't remember the state of it exactly, but I think it works.

[fleitz]

Looking pretty close to the real random and being actually random are two different things, which is why RNG analysis is so difficult.