Author: Sune Fischer
Date: 13:00:35 05/05/03
Go up one level in this thread
On May 05, 2003 at 15:32:43, Dieter Buerssner wrote: >On May 04, 2003 at 18:49:44, Sune Fischer wrote: > >>I am not convinced. >>I thought a long and hard about how to implement that imfamous draw KBP-K, where >>pawn(s) are on the rook file and bishop of the wrong color. > >There are many much easier problems. KQQK, KQRK, KQBK, KQNK, KQPK, KRRK, ... > >>The problem is not in detecting the specific case(s) as such, the problem is in >>detecting this along with possibly 30 other common patterns, and to do it >>quickly enough to not slow down the engine significantly. > >You can find some ideas in the papers I cited. Note that also the TB access >routines need to find the material pattern somehow before accessing the disk. > >>I think the best one can do is to build a kind of log2 pattern detector, if you >>know what I mean. > >You can do better, if you restrict yourself to the cases where TBs exist. One >smplistic example, have a table of function pointer, indexed by the material. >For example for kxyk, it could be a 5*5 table. In Heinz' paper a more >sophisticated method is explained. Okay, but then that leaves you with the problem Uri is worried about, I think. The typical "inaccurate" heuristics will tell you that three pawns are better than two, but with TBs the two pawn position might be evaluated exactly, thus to the engine it can appear as though two pawns are better than three under certain circumstances. This is the same phenomenon as when white refuses to take the pawn in KRB-KRP, because it can see that KRB-KR is drawn. To make a knowledge better than TBs you have to detect subpatterns within larger patterns, so that for instance in the above example white would understand that when KRB-KR is drawn, then KRB-KRP cannot be worse than a draw for black. -S. >Regards, >Dieter
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.