Levy, eh?

I have little faith in the assumption that stock returns obey some mathematically generated distribution ... like normal ... or lognormal or whatever. Them's mathematicians earning a living. Recently, R. Brown pointed out the Levy distribution. It (at least) has the efficacious property that LARGE returns are more frequent than, say, normally distributed returns. (That goes by the sexy name of fat tails.) I much prefer to use actual historical returns (over some sufficiently large time period) -- then randomly pick the returns from this collection. To put it bluntly: Use actual returns and NOT a mathematical proxy. However (I must admit) math models have fascinated me for years. What's particularly fascinating is to consider picking historical returns in a particular order - like from minimum return to maximum (or max to min). I did that once upon a time. The interesting thing is that the collection of historical returns has a particular distribution, a particular mean, a particular standard deviation, skew, kurtosis etc. etc. regardless of the order in which you select 'em.