Author: Jeff Lischer
Date: 14:07:53 12/19/01
Go up one level in this thread
On December 19, 2001 at 15:56:00, Roy Eassa wrote: >On December 18, 2001 at 23:17:40, Robert Hyatt wrote: > >>On December 18, 2001 at 20:15:09, Mike S. wrote: >> >>>On December 18, 2001 at 18:54:09, Kurt Utzinger wrote: >>> >>>>In the position shown below Fritz 7a sees that 1...Rxb2 is a fault, but after >>>>making this move, the program is unable to win a piece with 2.0-0-0 and so we >>>>must assume that there is a castling bug in Fritz7: >>> >>>I notice the same on my system. OTOH, Fritz 7a can see long castlings at the >>>root: >>> >>>[D]8/8/8/8/2p5/2pkp1N1/8/R3K3 w Q - 0 1 >>> >>>Analysis by Fritz 7: >>> >>>1.0-0-0# >>> +- (#1) Tiefe: 2/2 00:00:00 >>> +- (#1) Tiefe: 2/2 00:00:00 >>> >>>The case must be more complicated... >>> >>>Are there games where Fritz 7a did castle to the queenside (except book moves)? >>> >>>Regards, >>>M.Scheidl >> >> >>The bug is probably this: >> >>For castling, the king can't pass over a square that is attacked. On the >>kingside, Rxg2 would render castling impossible since the king ends up on >>g1 and is in check there. On Rxb2, it is very likely that the programmer >>simply tested one _extra_ square for being under attack (it is tempting to >>test all squares between the king and rook, which is right for kingside, >>wrong for queenside castling). >> >>I'd bet that is what he did. This bug has appeared in several programs. It >>was actually in a _very_ early version of Crafty as well... > > >But what causes Fritz to CORRECTLY see that ...Rxb2 loses to O-O-O in the >lookahead of the original position (Black to move), but NOT to see O-O-O after >...Rxb2 is played (and it's White's move)? Two different move generators, one >for lookahead and one for "now"? Or one for White and one for Black?? Can you ask Fritz to analyze the move O-O-O to see if it will analyze it when forced to? Maybe it's just a hole in the search when it's Black to move in this position.
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.