Quoting Brian Sheppard <sheppar...@aol.com>:
OK, so you don't have to worry if you set c==0. It might even be best. Just a note: in very preliminary experiments, c==0 is not best for Pebbles. If longer experiments confirm that, I presume it is because Pebbles runs on a very slow computer and searches only small trees. So your mileage may vary. But if c==0 tests well, then there is no reason not to use it.
I think it has to do with two things. Valkyria uses c==0 but I started doing that when I combined AMAF values with win rates. Even if only one move is searched AMAF values for nonsearched moves can go up and thus force the search to switch, even if the win rate for the most searched move does not go down.
Also with higher quality playouts you get a wider range of win rates for bad and good moves which lowers the need to search all moves as used to be the case before.
Other than that I agree that sticking to winning moves unless proved otherwise seems to work well. It just does not work fast enough if the search otherwise is ineffective.
Magnus _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
