Author: Robert Hyatt
Date: 14:58:14 11/08/01
Go up one level in this thread
On November 08, 2001 at 06:12:44, Mike Hood wrote: > >You can't make a chess program play worse by using EGTBs, can you? Not usually, >but look at the following study that involves castling. Crafty has all 5-piece >EGTBs at its disposal, in particular KRPKR. It's Black's move, Black has just >been checked by White's pawn, White can still castle. > >2k5/3P4/8/8/8/8/1r5p/R3K3 > >. . k . . . . . >. . . P . . . . >. . . . . . . . >. . . . . . . . >. . . . . . . . >. . . . . . . . >. r . . . . . p >R . . . K . . . > >Analysis by Crafty 18.10 (using EGTBs): > >1. -+ (-#17): 1...Kc8xd7 >2. -+ (-#18): 1...Kc8-d8 2.Ra1-a8+ Kd8xd7 >3. = (-0.01): 1...Kc8-c7 2.0-0-0 Rb2-b8 3.d7-d8Q+ Rb8xd8 >4. +- (#18): 1...Kc8-b7 2.0-0-0 Rb2-b6 3.d7-d8Q Rb6-c6+ 4.Kc1-b2 Rc6-b6+ >5. +- (#9): 1...Kc8-b8 2.d7-d8Q+ Kb8-b7 3.0-0-0 Rb2-b6 4.Rd1-d7+ Kb7-a6 > >Crafty takes the Pawn, not seeing that White can counter with 0-0-0+, and loses! >However, if Crafty is used with no EGTBs, the analysis looks completely >different: > >Analysis by Crafty 18.10 (no EGTBs): > >1. -+ (-7.20): 1...Kc8-d8 2.Ra1-a8+ Kd8xd7 3.Ra8-h8 h2-h1Q+ 4.Rh8xh1 Rb2-b1+ >2. +- (7.05): 1...Kc8xd7 2.0-0-0+ Kd7-e6 3.Kc1xb2 Ke6-f5 4.Rd1-h1 Kf5-g4 >3. +- (7.05): 1...Kc8-c7 2.d7-d8Q+ Kc7xd8 3.0-0-0+ Kd8-e7 4.Kc1xb2 Ke7-e6 >4. +- (7.20): 1...Kc8-b7 2.0-0-0 h2-h1Q 3.Rd1xh1 Kb7-c7 4.Kc1xb2 Kc7xd7 >5. +- (#10): 1...Kc8-b8 2.d7-d8Q+ Kb8-b7 3.Qd8-a8+ Kb7-c7 4.Ra1-a7+ Kc7-b6 > >Crafty makes the correct move, Kc8-d8, and wins. > >I've tested this position with every chess program available to me. Hiarcs 7.32, >Junior 7 and Shredder 5.32 all make the wrong move when using tablebases. Only >Fritz 6 and Chess Tiger 14 don't fall into the trap. How do they avoid it? By not probing if castling is legal? That is a solution that is easy to add. But the tiny cost is absolutely wasted as no 5 piece position with castling still possible will ever occur in any real game...
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.