by txandi » Wed Feb 24, 2010 11:15 pm
My two cents:
1) I think that the final map set should not be published. This makes the contest a real AI contest (no hardcoded solutions). When you design a system you usually need a test set and a validation set, so it should be the same here.
2) I also think that it would be a good idea to run a simpler tiny contest before the real one, with pairings consistent with the position on the leaderboard, in order to make it possible for all of us to improve our bots (I haven't done any important improvement of my bot during the past two or three days because I really don't know what is its behaviour against top bots).