Author: Robert Hyatt
Date: 08:30:38 03/30/01
Go up one level in this thread
On March 30, 2001 at 03:19:53, Uri Blass wrote: >On March 30, 2001 at 01:43:57, Christophe Theron wrote: > >>On March 29, 2001 at 14:21:14, Robert Hyatt wrote: >> >>>On March 29, 2001 at 14:08:06, Christophe Theron wrote: >>> >>>>On March 29, 2001 at 08:50:14, José Antônio Fabiano Mendes wrote: >>>> >>>>> An interesting miniature by a Russian composer(sorry,I can't remember >>>>> his name) I came across the other day: >>>>> [D]2N5/8/8/8/7p/8/7B/4K2k w >>>>> The _only_ winning move is 1.Bb8!! [mate in 34] >>>> >>>> >>>>Do I have a bug? >>>> >>>>Chess Tiger 14.0 and Gambit Tiger 2.0 both announce mate in 28 (that means they >>>>think it is a mate in 29 starting from your position). >>>> >>>>The line is: >>>> >>>>00:32:36.2 Mate in 28 20 354603928 Bb8 Kg2 Ne7 h3 Bh2 Kf3 Ng6 Kg4 Kf2 Kf5 Nf4 >>>>Ke4 Kg3 Kd4 Kxh3 >>>> >>>> >>>>I do not have all the tablebases installed, and it is why Tiger needed half an >>>>hour on my K6-2 450 to find the mate (24Mb hash). >>>> >>>> >>>> Christophe >>> >>> >>>Your PV is definitely wrong. The move Ng6 above is not optimal. Kd2 is mate in >>>30, after Ng6 it is mate in 31. Kf2 after Kg4 is also non-optimal and takes you >>>back to a mate in 31 for the 3rd time. After Kxh3 above it is a mate in 28 for >>>white from that point forward. >>> >>> >>> >>>However, it is possible this is an artifact of not having the right databases. >>>ie the P might turn into a knight in some variations and give check, rather than >>>just into a Q which might be worse. >>> >>>with all 5 piece files I get mate in 34 as I posted in a different post. >> >> >>OK, thanks. >> >>Indeed, I do not have all 5 men TBs installed. I have just installed the first >>CD (1/2) of the Endgame Turbo product from ChessBase. >> >>Sarah has posted in another message that Tiger announces instantly a mate in 34 >>with all TBs installed. No surprise as the root position has 5 pieces. :) >> >> >> Christophe > >It is clear that there something is wrong here. > >There are two possibilities > >a)Tablebases are right and Tiger is wrong. >b)Tiger is right and tablebases are wrong. > >I have nothing against Tiger but For some reason I guess that a is correct. > >Playing a game between Tiger as white and Tiger with all the relevant tablebases >with black can give us more information. > >Uri One problem is promotions. If he doesn't have a promotion table for that pawn, then the search may well find positions where it can force a promotion at the right instant which _seems_ to avoid getting mated. If it had the right table, it would see that the promotion doesn't do a thing to avoid the mate, actually. That would be my guess. however, announcing a _shorter_ mate than 34 is odd. I never see "shorter" mates announced. I often see longer than expected mate announcements because the program simply can't see the most efficient path to a tablebase mate all the time. So I'm with you to an extent here. If it is a mate in 34, and a program says mate in 30, that would be cause for concern. It might not be a problem, due to the promotion stuff, but it is strange.
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.