Author: Jaime Benito de Valle Ruiz
Date: 09:30:56 05/13/03
Go up one level in this thread
On May 13, 2003 at 00:56:49, Jouni Uski wrote: >[D]5bk1/p3p1p1/p3P1P1/p7/p7/8/P7/5K2 w - - > >This is one of my oldest test positions with mate in 17 (Ke2). It's still >unsolved by top pros, but Yace and Crafty can solve it. Interestingly >Shredder7.04 first says mate in 22 and then mate in 0 (#0) !?!? > >Jouni This is another "highly improbable" position where the moves of one of the sides are only a few (if not one) forced moves, and at the same time, it has a considerable material advantage. The problem with this type of positions seems to be that the Null-Move (among other prunning-methods) is causing more damage in the search than anything else, because it prunes all options due to the fact that one side has a considerable material advantage and no apparent improvement takes place for anyone. The search finds the repetition draw sooner or later, and any other options are discarded by the null-move because it doesn't "see" the mate soon enough, and one is loosing from the material point of view. This is just a guess, of course. Another plausible theory could be that some programs decide to stop the search if the score has not improved significantly after several ply, and the other side is clearly better (again, from the material point of view). Maybe any other similar kind of reduction, such as this-seems-to-be-a-draw reduction; who knows? Fritz 8, for example, doesn't find anything but a plain draw firstly, and then, gradually and after several seconds, the score increases, but still no mate. Change the Selectivity from 2 to 0, and it will find one mate in no time (#24). Untick the Futility Prunning, and it will announce #18 in no time (Hash cleared every time). I remember that my program could not find one of these obvious forced mates when I first tried. Then I turned off the Null-move, and solve the problem instantaneously. That version had still no reductions or futility prunning yet, though. In any case, it is interesting to see how these positions are easily solved by just adding some more "useless trapped pieces" to balance the material score: After adding some pawns and 2 bishops, some programs like Fritz 8 or Schredder 7 find "a solution" (not necessarily the best one) fairly quickly: [D]5bk1/p3p1p1/p3P1P1/p3P1PB/p5PB/6P1/P7/5K2 w - - 0 1 Regards
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.