Author: David Blackman
Date: 21:39:31 12/02/99
Go up one level in this thread
On December 01, 1999 at 21:34:57, KarinsDad wrote: >On December 01, 1999 at 19:40:28, Jeremiah Penery wrote: > >>While watching Crafty play some games today, I thought of an interesting, but >>likely unfeasable idea. The problem was that one of Crafty's pieces was >>hopelessly trapped, but it pushed the loss over the horizon. It still lost the >>piece. >> >>The idea is a way of detecting such trapped pieces, and evaluating them >>properly, so they will have no chance to be pushed over the horizon. >>Do a small search with each piece (not pawns) at the root - at least a 2-ply >>search on all the moves for this piece. If the search finds that the piece >>can't escape, then the score drops. The problem is that this probably won't do >>much good once it happens at the root, and it may be unfeasable in the tree. >> >>However it's done, there should be a way of detecting trapped pieces. :) > >I have plans to do this, but I haven't yet come up with a good idea on how. It's >an extremely non-trivia problem. > >KarinsDad :) KnightCap has trapped piece code that works quite well. But it is slow to calculate. When KnightCap beats Crafty (which is not all that common) it is usually because KnightCap traps one of Crafty's pieces and Crafty doesn't even notice until it's too late. It didn't sound too convincing when i heard the method explained, but it seems to work ok in practice. It is purely static eval, see the source.
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.