On Mon, 11 Oct 2010, terry mcintyre wrote:
From: Thomas Wolf <[email protected]> > I am running a test with a player of professional strength to guess the next > move in professional games, all after 100 moves have been played. After > running about 200 positions the average is about 4.4 guesses. This does not > tell the distance to optimal play but it tells something. I do not understand. Does this mean that the "player of professional strength" guessed only 4.4 of the 200 moves?
No.
Or does it mean that the player required an average of 4.4 guesses per position to find the next move?
Yes.
I'd be curious about the distribution.
We get this too. Btw. you and everyone is welcome to participate: http://lie.math.brocku.ca/aj07ll/ The reason why I am doing this is to get a calibration of the frequently used test of static evaluation functions (mainly pattern matching programs which learned pattern from pro games) to guess the next move in professional games.
As Pietr remarked, pros often find more than one good move, and its hard to say which of the candidates is the "best" move of the lot.
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
