Now, I'm more into my stats than most people, I know that. So for my upcoming random switcher hack, I decided to check just how random it was. I made sure that the hack generated some loggable messages I could parse, ran with -condebug, generated 10,000 random events and loaded them into
R.
The results were disappointing. Although the number of "Choice A" and "Choice B" generated were nearly identical, the chance of getting "AA" (an A followed by an A) or "BB" was higher than getting "AB" or "BA" (all four should be 25% probability if the system was fair).
I had spent several nights trying to fix this with various reconfigurations of the entities, but never managed to make significant headway. Then I decided to knock up a quick mod which would spit out A and B into the console using the random() function directly, just to check. And of course it turns out to have this same issue but
worse. So I will be releasing this map hack shortly...