Author: Colin Frayn
Date: 03:51:46 11/23/01
Go up one level in this thread
On November 22, 2001 at 22:22:13, Christophe Theron wrote: [D]8/1KPk4/8/6p1/p7/8/8/8 b - - >A simple definition of "quiet" is that a position is not quiet if one side is in >check, one side has a winning capture, or one side can promote. Well, firstly, it's black to move. Secondly, one could easily retract the last couple of moves, getting to this position, [D]8/8/1KPk4/6p1/p7/8/8/8 w - - ... which is also obviously won for white, but which nobody spots statically. You're now quite a few ply away from the winning promotion. Fortunately I changed Beowulf's eval() so that it now spots similar positions to these quite easily. The reason why programs don't understand them is the following; (and Crafty scores this as something like +7.5 pawns for black) Black has two totally passed pawns (safe from opponent king, no opposing pieces) which are scored extremely highly. White has one passed pawn, but it is not even safe from the black king. Well... in fact it *is* because it can be shown that white will always be able to promote this type of passed pawn (assuming no other opposition) because his own king is defending. Anyway, I only conssider pawn races for totally passed pawns, so the first important change is to recognise that this white pawn is in fact totally passed. Secondly - the fact that white will promote first effectively means that black's two passed pawns are useless - there's no point giving them the extra bonus for being totally passed as soon they won't be - i.e. there will soon be an opposing queen to deal with, and they'll both be captured. If we don't give them huge bonuses (in Beowulf, that bonus is 4 pawns worth) then the situation is scored as a strong win for white. I keep finding these types of problems. And no, the answer isn;t to use tablebases, as it's reasonably easy to construct a situation with a few extra pawns in where the same rule applies. Cheers, Col
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.