Thanks for the responses, I will try attaching the various versions to CGOS. To answer a specific question
> What board size, and how many playouts per move? > 9x9. 10k gives about a 5% winrate as black, 50k maybe 17%, and 100k a solid 20%, but it's not so much the winrate as the fact that the lost games are almost all W+87.5. - James On Tue, Oct 19, 2010 at 1:50 PM, Christoph Birk < [email protected]> wrote: > On Mon, 18 Oct 2010, James Pettit wrote: > >> I have a basic UCT bot with AMAF/RAVE implemented as part of my master's >> research. The bot can muster a small win-rate versus Gnugo level 6 (~20% >> as >> black), but usually loses games by not forming 2 eyes and being completely >> captured. I have uniform random playouts, with simple ko and self-eye >> moves >> disallowed. Is this normal for a basic UCT bot? >> > > I recommend running it on CGOS (9xx), first with UCT + AMAF turned off. > Compare it to > myCtest-10k > myCtest-50k > They do 10000/50000 random playout (no UCT nor AMAF) > > Then turn on UCT and compare to > myCtest-10k-UCT > myCtest-50k-UCT > > I also have two programs that do AMAF only (no UCT): > myCtest-2k-AMAF > myCtest-10k-AMAF > > These programms are running for quite a while and should > have a stable rating. > > Christoph > > > > > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
