Author: Robert Hyatt
Date: 12:34:05 09/01/99
Go up one level in this thread
On September 01, 1999 at 14:15:11, blass uri wrote: >On September 01, 1999 at 11:28:41, Robert Hyatt wrote: > >>On September 01, 1999 at 04:55:34, blass uri wrote: >> >>>r1b1k2r/1pqp1ppp/p1Nbp3/8/4P3/2N5/PPP2PPP/R1BQR1K1 b kq - 0 1 >>>avoid Bxh2+ >>> >>>Junior5 solved this test againsy yudasin at tournament time control but could >>>not solve it at faster time control >>> >> >> >>Here is my output for this. Crafty doesn't think it is 'good' at all, but >>does bounce between that move and dxc6/bxc6 as it searches. By the 1:33 >>mark it has decided that this is getting bad enough that it discounts it >>and changes for good at depth=13... >> >>It finally (after finishing depth=13) settles on bxc6. If it has a target >>time of 94 seconds (or longer) it won't play Bxh2. In looking at it, Bxh2 >>doesn't look horrible, since white has castled and black has not, although >>I doubt I would play it myself. > >GM Yudasin gave Junior the opportunity to play Bxh2+ because yudasin did not >think that Bxh2+ is a good move >> >> >> >> >>>r2q1rk1/pb3ppp/2n5/1pbNP3/2p1Q3/P4N2/BP3PPP/R4RK1 w - - 0 1 >>>from Junior-nimzo(wccc) >>>find Nf6+ >>> >> >>This is a tactical shot that really has nothing to do with king safety. Unless >>you weigh king exposure _way_ high, which will find this move, but will also >>find lots of sacrifices that are totally worthless. > >I do not know about programs that can see in a reasonable time that white is >winning more than 3 pawns for a piece so the question if a program can solve it >after some minutes is a question of the evaluation of king's safety. > >sometimes 3 pawns and exposed king for a piece is a good idea and sometimes it >is not a good idea and a program with good king safety evaluation should know >when it is a good idea and when it is not a good idea. > >Uri the first question is, "is it good in this position?" I only know of one program that does the kind of analysis necessary to answer this reliably... Deep Blue... because they can (in hardware) detect which pieces are attacking the opponent's king field from a distance (sliding pieces), which pieces are lined up in battery on the king, etc. I could do some of this in Cray Blitz because loading the vector (diagonal/file) took no longer than loading one square on the ray... but in Crafty this would _really_ be too slow, probably, until the hardware speeds up enough to avoid dropping down to 10-11 plies in the middlegame... But it is an interesting question... and it let us see some things in CB that Crafty simply doesn't have a grasp of...
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.