For a while now I've been interested in sports / games 'ratings' systems. Some years back I collected results from weekly football games and wanted a way to score individual players from team score results (the teams change each week) - ultimately I wanted to help pick balanced teams). This year I set up a company table tennis ladder on racquetladder.com but was desperately unhappy with the ranking system - the winner / loser simply switch (or stick) ranks, which is terribly unstable (although very transparent). So I made a python / numpy implementation of Rémi Coulom's 'Whole History Rating' algorithm, and experimented a little with creating 'interactive' SVG plots of the results.
Then I took part in the Planetwars "Google" AI challenge, and chose to use Python, thinking it would be dominated by clever heuristics rather than brute calculation. Top UK Python entry (*bow*) but, sadly, other languages dominated. Interestingly, they used Coulom's 'Bayeselo' to rate the entries, about which there was much griping, but nobody could think of a better way to do it -- I think generally people underestimated just how powerful Bayeselo is, and besides, those people were all over on kaggle.com where they concurrently had a 'design a better Elo' competition! So, for me it was an interestingly coincidental year. Oh, I also went to a London Dojo -- it was a lot of fun and I hope to go again. I don't really have any projects on the go at the moment so 2011 is a blank slate. Definitely want to win the next ai challenge :) Nice to have a chance to talk about these things. Matt _______________________________________________ python-uk mailing list python-uk@python.org http://mail.python.org/mailman/listinfo/python-uk
