Quoting Matt Gokey <[EMAIL PROTECTED]>:
Magnus Persson wrote:
Quoting Matt Gokey <[EMAIL PROTECTED]>:
I was wondering if anyone was combining an opening library with MC/UCT
by running a large number of the simulations and storing the results.
This seems like a pretty natural extension to save simulation time in
the opening.
I actually first tried to save the entire search tree for every position. Then
it would load the position and search deeper every time it was
encountered. The
problem was simply disk storage, and there is always positions where
the program
itself cannot in a reasonable time find strong moves by searching.
How strong a player are you? You are probably unfairly evaluating
Valkyria based on your strong expert play/perspective. I'm rather
amazed that MC simulations find good moves at all given that most of the
playouts are nonsense games. That is why I say MC/UCT is finding
reasonably robust moves with more favorable options (strategic play) not
necessarily great/best moves. Because of mostly meaningless playouts it
misses nuances and tactics deeper into the game that would show
otherwise. It seems the deeper the simulations the worse this effect
would be.
I am european 2Dan. I agree that MC/UCT finds robust moves, it is more
important
to avoid mistakes in go rather than finding the best move. There often little
difference between alternative "bestlooking moves" but a blunders can be huge.
I still disagree though a little on "misses nuances and tactics". I
think there
is always a "signal" from the simulations that MC can pick up to some degree.
The signals that get lost or cannot be distinguished are discoverd by
searching
deeper with UCT.
-Magnus
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
