Author: Robert Hyatt
Date: 15:26:01 12/08/00
Go up one level in this thread
On December 08, 2000 at 17:07:10, Uri Blass wrote: >On December 08, 2000 at 16:52:46, Robert Hyatt wrote: > >>On December 08, 2000 at 15:25:06, Torstein Hall wrote: >> >>>On December 08, 2000 at 14:47:37, Dann Corbit wrote: >>> >>>>[D]2r3k1/5p2/1p2pp1p/pPq5/Pn6/1B6/1P2QPPP/3R2K1 w - - 0 25 >>>> >>>>It seems some programs get the right answer immediately (e.g. Phalanx) and >>>>others simply never find it. >>>> >>>>My question is, "Why?" >>>> >>>>I tried programs that don't do NULL move, and that did not seem to help. So >>>>what is preventing programs from finding the right choice? >>> >>>...and Bxe6 is the right move I guess!? >>> >>>Found by Gambit Tiger in a split second >>> >>>Torstein >> >> >>You can find this one of two ways. You give up a piece for potentially three >>pawns + no king safety for black. If this were all there is, it would be a >>mistake as white is a piece down. But there is more. Going deep enough, black >>has to give the piece back, and white simply ends up winning a pawn, period. >> >>It is not hard to make any program like this. Crafty hates the idea of giving >>a piece for three pawns because 90% of the time that loses. > >I agree that a piece is better than 3 pawns in most of the cases but >I think that it is an exagaration to say that in 90% of the cases the knight is >winning. >There are many cases when knight for 3 pawns is a draw. Depends on where it happens. In the middlegame, I think it loses, period. In the endgame, the three pawns can draw, and on rare occasions win. But in any case, I will rephrase... when I added the "bad trade" code to crafty, it was done after analyzing about 100 games played on ICC by Crafty. And in those games, about 90% of the time the side sacrificing the piece for three pawns lost. Whether that would be true now, I don't know. > > It has to see the >>"meat" before it will "bite". Which takes 14 plies and 3-4 minutes on my xeon. >>If a program has big scores for king safety, it will probably bite on this >>instantly. If not, it will have search deep enough to see that black has to >>sac the knight to avoid getting mated, leaving white a pawn up in what might >>still be a drawn ending (both sides have pawns, plus rook and queen each. > >The main question is if white can win after Bxe6 fxe6 Qxe6+ Kh8 > >White can take 4 pawns for the piece but I do not see a forced win for white and >I suspect that it is a draw. > >Black does not have to sac the knight in this line but only to sac a pawn. > >Uri Very possibly. I am not sure either line is actually winning.
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.