I think it's a valid method.

I load Amit's prediction list into memory. I reset all predictions to 6.5 magnitude. Then on each random run, all I do is quickly run through all the predictions and randomize the date. I make sure the range of dates is with the same range as the original list.

Anyway, I got it done and this method is orders of magnitude faster. Dunno why I didn't think of it. My run of 10,000 was done in about 18 seconds. Got the same exact graph because I used the same specific seed value for the random number generator.

Brian