I do doubt that it would make engine any stronger but please do try if you like. I once tried something like that but did not help the engine at all. The GG evaluation is not really suoted to tree search. It gives values to moves not to boards situations. And good evaluation function for go are hard to come by.
Basically all good programs now are Monte Carlo Tree Search programs. Solves the evaluation problem nicely. So if someone is still interested on improving gnugo the only thing to do would be trying out using gnugo knowledge to bias MCTS. You could experiment with pachi using GG to give a prior for search or something like that. Petri 2013/5/15 Xi Tao <xta...@students.poly.edu> > Dear Gnu Go, > > My name is Xi Tao, but everyone calls me Tony. I am interested in making > the alpha beta reader that was mentioned in the task list in order to make > predictions one or two moves deep to prevent fatal errors. I have > experience in using alpha beta search, and have created a game engine for a > 4X5 version of the connect four game with the drop out rule. Thanks! > > Sincerely, > > XI (Tony) Tao **** > > _______________________________________________ > gnugo-devel mailing list > gnugo-devel@gnu.org > https://lists.gnu.org/mailman/listinfo/gnugo-devel > >
_______________________________________________ gnugo-devel mailing list gnugo-devel@gnu.org https://lists.gnu.org/mailman/listinfo/gnugo-devel
