I was not at all surprised by this result.
My thinking goes like this. On 9x9 the global situation is everything
that matters and precomputed information is not as important as
searching effectly is. Good 9x9 games are often very sharp fights
where then next move often violates good shapes known from 19x19 games.
On 19x19 deep global search is impossible and good local shape is more
often more important (or perhaps more often local shape is good enough
to indicate the correct move) than the global situation so heuristic
precomputed knowledge is essential to make search efficient.
In Valkyria I also learned that things I do to improve 19x19 has no
effect on 9x9 or sometimes makes it weaker.
-Magnus
Quoting David Fotland <fotl...@smart-games.com>:
So far on 9x9 go, Many Faces doesn't seem to make a huge difference. On
19x19 it makes a huge difference. I ran test games overnight against Gnugo.
With Many Faces turned on, my engine wins 85%. With many Faces turned off,
my engine wins 7%.
Both results are unexpected. Since most of my tuning is on 19x19, the
combination of Many Faces and UCT is probably badly mistuned for 9x9.
David
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
