I just modified my bot to do just playouts, no tree, move selection is most wins. It's on CGOS now as GoGo-10k. I'll leave it there for a while. It definitely exhibited some of the same problems as the UCT version, but I cannot for the life of me find a bug in my playout implementation. - James
On Wed, Oct 20, 2010 at 4:24 PM, Christoph Birk < [email protected]> wrote: > On Thu, 21 Oct 2010, Matthew Woodcraft wrote: > >> (Unlike what others are reporting, I do see a big improvement from RAVE >> even with uniform random playouts.) >> > > I think we agree. I see a big improvement with AMAF (not RAVE) > with uniform playouts (no UCT) too. > But once I use UCT the improvement of AMAF (not RAVE) is only > minimal. > I use the terms > AMAF: all move as first (after the simulations) > RAVE: use the AMAF values to seed the (UCT) tree search > > I would like to repeat this suggestion: > Run your uniform random bot (no AMAF/UCT/RAVE) on 9x9 CGOS > with 10k or 50k playouts and compare to myCtest-10k/50k. > This should tell you if you ave a (fairly) bugfree implementation > of the "don't-fill-your-own-eye" rule. > > 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
