Can you forward a copy to the Empeg guys? I wanna know why a full-player-shuffle still tends to cluster neighboring FIDs.
Ha! I'm not statistician, this document is about a very specific program and I doubt it would work to cover anything else. It basically states (in a much longer format) that given a small number of tests pure randomness will tend to favor certain values and yield a bell curve, but over a larger number of tests this curve will flatten. However, as we tend to perceive things in smaller sections we see the outliers on the bell curve and assume the program isn't random. In order to prove that the program is working I actually broke the selection code out into a separate program where I could simulate 100,000 tests and show that the curve is flat (all items are selected at an equal percentage) over extended use. However, any given set of 50 tests will appear to have items "favored" over others.
Anyway, my document is much longer and more detailed, but that's the basic jist. It really just explains why my test proves that the program is truly random, even though it isn't perceived that way. I’ve had calls daily on the subject, and trying to recall all my statistics from college has been making my head hurt. At one point I even plugged a binomial distribution function into a spreadsheet in order to give users specific expectations based on their data, and this shuts most calls up (mostly because they don’t understand what I’m sending them, but it works). This document is being prepared for the head bigwig, who needs more than just a simple Excel spreadsheet.