Talk:World 3/@comment-27452760-20170325013421/@comment-26082284-20170327190447

Actually, we don't know how exactly Kancolle rng works. Whatever it is - on-demand algorithm or pre-generated tables - those aren't actual coinflips. And while we can safely say that it's generally fair - that is in very long runs you get approximately equal numbers of 'heads' and 'tails', I'm not so sure about how random distribution of results is. I can't shake the feeling that on some maps rng tends to have sort of hangs up when it flips you one way whole day, and then next day it flips you the other way, so in the long run it's equal, but short term... not at all. 3-3 is one of those maps. Also 2-2, though this one is harder to pin down due to routing affected by composition. Others, like 4-2 or 5-5 seems to be more random in comparison.

One day I'll get and start doing long-term statistic to prove it one way or the other.