Welcome Jonas. [EMAIL PROTECTED]: <[EMAIL PROTECTED]>: >> I experimented with something similar a while ago, using the >> publicly available mogo and manipulating komi between moves. >> >> If its win probability fell below a certain threshold (and the move >> number wasn't too high), I told it to play on the assumption that it >> would receive a few points more komi (and similarly when the win >> probability became high). > >That's certainly a nice solution ! > >It's probably easy to implement, and relatively easy to tune.
I've implemented and used the idea for the first UEC Cup Computer-Go tournament (19x19, Japanese rules) on Dec. 1-2 and was 5th/27 participants. http://jsb.cs.uec.ac.jp/~igo/eng/ #No results in English pages. It's easy to implement but hard to tune, though by my experiments. The point is how to caliculate an appropriate bias or komi from current info, ie,expected winnig rate, number of moves, etc. I've tested some means and used following very ad-hoc formula for the tournament. delta_komi = 10^(K * (number_of_empty_points / 400 - 1)), where K is 1 if winnig and is 2 if loosing. Also, if expected winning rate is around 50%, Komi is unmodified. By a few tens of benchmark games against GNU Go before the tounament, it improved about 100 ELO but the number of games is not enough to believe this. >Now, the question is: did it seem stronger to our eyes because it played >more human-like ? Yes, certainly. I've tried and posted the use of tanh(K * score) to this list on Dec 13, as well. Hideki Kato: <[EMAIL PROTECTED]>: >Hi Begué and Don, > >I did this in my earlier version of ggmc. The real code was: > >reward = 0.5 * (1 + tanhf(K * (score - komi))); ># tanhf() is a float, not double, version of hyperbolic tangent >function. ># I use tanh() as exp() may cause overflow. ># You can see the code from http://www.gggo.jp/ as it's still left. > >I got best performance around 10 of K but it's so little that I'm >using simpler one now. -Hideki >As a remark, tests of strength with this kind of strategy could be made >through self-play with handicap games. > >Moreover, if a program plays >better when he is ahead (or conversely behind), it might be an idea to >make the program play initially with a higher (resp. lower) komi. >I seem to remember someone saying that a MC program (I don't remember which >one) was better when he had handicap stones, than against them, because >he used these stones well. I wonder if there was any link with the >estimation of position. > >Jonas > >_______________________________________________ >computer-go mailing list >computer-go@computer-go.org >http://www.computer-go.org/mailman/listinfo/computer-go/ -- [EMAIL PROTECTED] (Kato) _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
