by zeiris » Mon Feb 22, 2010 11:01 pm
There are at least two major problems facing the tournament right now:
-very variant ranking makes it impossible to compare bots closer than ~50 ranks apart
-huge amount of randombots increases variance of ranking and makes it harder to judge overall performance of real bots
Currently, the benefits of letting the contest run (the learning and teaching that result) outweigh the fairly empty benefit of randomly determining a "winner".
As well, from a more idealistic viewpoint, the point of tournaments like these is to hopefully advance technology and thinking, and maybe even discover some new techniques (or new ways of applying old ones!) With the current ranking system, we can't tell if that's happened. I realize everyone involved is doing so on their own time out of goodwill, but extending the tourney and using that time to improve the system to keep up with the improved bots seems like the best thing to do.