Yes, sort of. I've used it for calculating a resampled efficient investment frontier. The great strength of perl for this task, I think, is the ability to munge input data quickly and easily and to quickly recode your program to test variations. It may not be the fastest simulation, but the speed of development/coding makes up for the speed of running, at least for the purposes I have for it. That said, PDL is certainly the way to go if you're dealing with any large data sets. On the statistics front, I wound up rolling my own so I knew exactly what was going on, so I'm not sure of a good recommendation there. (This from my early perl days -- now I'd probably look for a good module, or else would write my code in a way I could then release. It's on my "one of those days" list to revamp and release it.)

I'm in the process of coding some other simulation stuff right now -- keep your eyes peeled on CPAN in the next month and you'll probably see it. It may or may not be useful to you, but I plan to include support for bootstrap resampling, etc., which is nice for backtesting investment returns.