Author: Stefan Meyer-Kahlen
Date: 07:19:06 02/08/00
Go up one level in this thread
On February 07, 2000 at 16:10:55, Ernst Walet wrote: >I´ve seen some strange evaluation glitch in Shredder 4 with the following >position. > >8/1K6/8/5k2/bR1P1r2/1n6/8/8 w - - > >Shredder wants to play Rxa4, score ply 9-14 about -1.2 > >Make the move. > >Shredder wants to play Nxd4, score ply 9-14 about +4.71 > >Make the move. > >Shredder wants to play Ra8, score ply 9-14 about -1.3 > > >Why such a different score after the first move? Does Shredder much >preprocessing? > >Ernst-J. Thanks for your interesting post! First of all, Shredder's evaluation is always from the point of view of the moving side, so there is "only" a difference of 3.5 pawns and not 5 pawns :-) This is a special case for which I added code after Jakarta 96, where Shredder almost blew a game against Virtual Chess. I, like many other programmers, evaluate a KRN-KR ending very drawish. In your second example, I simple deactived that code so Shredder as black this time will go for the last white pawn. If the code were still active, Shredder would never capture the pawn, because KRN-KRP would be better than KRN-KR. I agree that there are other ways to solve that issue, but I haven't touched that code for quite a while now and almost forgot that this is still in there. I agree that this is very brute force, but that method proved to be very succesful in chess :-) Hope this helps Stefan
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.