Author: David Rasmussen
Date: 11:04:47 06/12/01
Go up one level in this thread
On June 12, 2001 at 11:45:00, Daniel Clausen wrote: >Hi > >On June 12, 2001 at 10:05:50, David Rasmussen wrote: >[snip] >>I see all your points (in fact, I saw them before you made them :), but I >>_still_ think it must be nicer to somehow have transpositions from non-book >>positions to book positions and simultaneously avoiding the problem you >>mention, if it can be done. > >Well, what you can do is this: your search prefers moves which lead to >book-positions, as long as there's no other moves which have a score higher than >the move which leads to a book-position plus some margin. > >Example: > >Searching from non-book position and come up with 2 moves > >Move A has score +0.15 which leads to a book position >Move B has score +0.23 which leads to a non-book position > >If your margin is below 0.08, you'd choose move B. If margin is bigger (or >equal, sheesh :) you'd choose move A. Crafty would choose move B because it uses >a margin of 0.0. > >What would be a reasonable margin is another question. You could make this >margin fixed, dynamic (depending on position and whatnot) etc. An engine with a >sophisticated eval would prolly choose a lower margin. > I've been thinking of a such an approach myself, but there will still be a problem in the very cases where opening books are important, in an opening like QGA, where black should not try to keep the pawn, but search is not likely to reveal that. So if the program is black, and is contemplating b5 to guard his c4 pawn, it will see that this move wins a pawn. So the opening book says that b5 is not a book move, and maybe Nf6 or something is, then it would require a large margin of maybe one pawn to make it play the book move. This would be good in this case, but a margin of 1 pawn might be too much in other situations. Of course, the position after b5 will be in a good book, and denotes as a bad position for black, which black never enter. But then it requires that the book has no holes of "bad" positions, which is impossible. >Personally I like the way Crafty handles it. (because of KISS) > I like it too, I just can't shake the idea that there might be something better.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.