Author: Mike Hood
Date: 10:29:38 02/18/04
Go up one level in this thread
On February 18, 2004 at 12:26:39, Ingo Bauer wrote: >Hi > >>>[D] 4k3/8/8/8/8/8/3q4/4K3 w - - 0 1 > >>While Shredder 8 is probably the strongest engine currently available, there is >>no question it is the buggiest. I've seen more Shredder bug reports to this >>forum than all other engines combined. > >Really, do you have a collection of REAL bugs, or is it just the case that a lot >of users post something where they THINK it is a bug? > >This example with 5 pcs Tbs installed brings in the Classic GUI: > >Engine: Shredder 8 (256 MB) >by Stefan Meyer-Kahlen > 4/04 0:00 0.00 1.Kxd2 (3) 0 >best move: Ke1xd2 time: 0:00.016 min n/s: 250 CPU 93.7% n/s(1CPU): 266 >nodes: 4 TB: 3 > >seems quite ok > >and with the CB GUI i get: > >Neue Partie >4k3/8/8/8/8/8/3q4/4K3 w - - 0 1 > >Analysis by Shredder 8: > >1.Kxd2 > = (0.00) Tiefe: 1/1 00:00:00 > = (0.00) Tiefe: 4/4 00:00:00 > >(, MyTown 18.02.2004) > >which seems ok too. > >So I suspect that there is something wrong with the original posters setup >(TBS?) > >Bye >Ingo Bauer I can't check if Shredder works correctly in the classic interface, because I don't have it installed on my PC at the moment. But I can assure you that there is a bug in the Chessbase GUI. Check my original post again, Ingo. In the above position there are two legal moves, Kxd2 and Kf1. Shredder 8 considers BOTH of these moves to be a draw. Set your infinite analysis to display several lines in parallel, and you'll get this result: Analysis by Shredder 8: 1. = (0.00): 1.Kxd2 2. = (0.00): 1.Kf1 Kf7 It seems to be a combination of two bugs. The first bug is that the KQ-K tablebase is not accessed, even though it exists. (There are utilities available that monitor file accesses, and I verified that no tablebases are accessed in this position). The second bug is that Shredder 8 ceases to calculate, even though the analysis ought to be infinite. "White moves his king to f1. Black moves his king to f7. Nobody has won yet. Draw!" A good engine should be able to find a Mate for Black in this position even without tablebases. Look what Crafty says: Analysis by Crafty 19.08: 1. = (0.00): 1.Kxd2 Kd7 2.Ke3 Kc6 3.Kf4 Kb5 4.Kg5 Ka4 5.Kh6 Ka3 6.Kh7 Ka2 7.Kh8 Ka1 8.Kg8 Kb1 9.Kh8 Ka1 2. -+ (-#6): 1.Kf1 Crafty searches to depth 60 to come to this result. Shredder only searches to depth 4, doesn't find a mate, and stops. Concerning the first bug, I can guess what is algorithmically wrong with the tablebase engine. It always assumes that there is a smaller tablebase available after a capture, and if there isn't it switches itself off and uses the main engine. This is sensible, for instance if a user has the 5-piece tablebases but has forgotten to install the 4-piece tablebases. But in this case, the KR-K tablebase is available, but there isn't a K-K tablebase, so it's the wrong decision.
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.