> If instead of assuming a player's performance follows a symmetrical > probability distribution function (the normal in the case of Elo), > assume the p.d.f is asymmetric, e.g. triangular. If player's have > p.d.fs which are increasing, decreasing or rectangular, wouldn't > that produce intransitivity? I don't know if it is the case, but simulation > should make it clear. If this is the case, the advantage is it can be created > only from the results of matches and not judging a player's ability for this > and that.
Even if each player's performance is asymmetrical but identical, the difference of performance becomes symmetrical again. But still, intransitivity can be seen from results of matches. If one has enough results of N people playing against each other, one could use the vector of performances against each opponent as an input to some machine learning method, such as a neural network or principal component analysis. I would assume that the first principal component would represent strength and the second would give some kind of intransitivity. If someone has result data with dense enough pairings, I could run some experiments. BTW, I also have some ideas on how to make UCT work on the big board. They are partly published in a paper using games of Y and Hex as examples (it did not use UCT, though). T. Raiko. Higher Order Statistics in Play-out Analysis. In the proceedings of the Scandinavian Conference on Artificial intelligence, SCAI 2006, pp. 189-195, Espoo, Finland, October 25-27, 2006. http://www.cis.hut.fi/praiko/papers/scai06.pdf Tapani -- Tapani Raiko, <[EMAIL PROTECTED]>, +358 50 5225750 http://www.cis.hut.fi/praiko/ _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
