Author: Uri Blass
Date: 10:11:33 07/23/01
Go up one level in this thread
On July 23, 2001 at 11:50:31, Robert Hyatt wrote: >On July 22, 2001 at 09:02:52, Sune Larsson wrote: > >>On July 22, 2001 at 00:18:39, Robert Hyatt wrote: >> >>>On July 21, 2001 at 21:56:16, Sune Larsson wrote: >>> >>>> >>>> 6k1/6pp/3n4/1n1p1P2/p1pP1NP1/PpP1K3/1B5P/8 b - - 0 1 >>>> >>>> Euwe - Averbakh, Zuerich 1953 >>>> >>>> >>>> In this position white has just played 36.Nh5-f4, casting eyes on >>>> the squares d5 and e6. Averbakh, just like a Spanish matador, went >>>> for the swift kill, and just 4 moves later Euwe resigned. It happened: >>>> >>>> 36.-Nxa3! (clears the road for the pawns) 37.Bxa3 Nb5 38.Bc1 Nxc3 >>>> 39.Ne2 Nb1! and white resigned 0-1 >>>> >>>> >>>> It looks like Deep Fritz has some trouble with this position. >>>> DF wants to protect the pawn on d5 with 36.-Nc7, tried it so far >>>> as ply20 and 10 min on a PIII 800. >>>> >>>> >>>> Test: The mission is clear and the direct winning line above. >>>> >>>> Sune >>> >>> >>>This is easy with the right knowledge: >> >> >> Yes, and this position causes severe problems for Hiarcs732, Nimzo732/8, >> Fritz532, Fritz6 and Deep Fritz. They all want 36.-Nc7. >> >> Solved instantly by Tiger, Junior (even J5), Shredder (even S3) >> and CM (though some newly created personalities might need more time). >> Also Yace finds the correct continuation after less than 1 minute. >> >> Appearantly this knowledge is missing in F6/DF. For the latter I aborted >> the analyse after 10 min, though. >> >> Sune >> >> >> > > > >This is just a weakness you have exposed in Fritz. It has many. It is not >a weakness in Crafty. But it also has many _other_ weaknesses as well. > >The point here has to be that if it takes Fritz 3+ minutes to find this, then >it would not try to reach this position deep in the search because it wouldn't >know it is winning (nor would it try to avoid it because it sees it is losing). > >IE if you back up 1/2 move and have fritz search, it probably won't be able >to see the tactics then, and without the positional judgement to help it under- >stand what is going on, it might make the same move played in the game, allowing >the opponent to win simply. In this case the last move of white was not a blunder and white had nothing better. I can also add that finding the right move does not prove that the program has the right knowledge. Programs can find it by increasing the value of pawns when the same knowledge may be counterproductive in different positions. Fritz also has no problem to see that Nxa3 gives advantage for black and the main problem is that it believes that Nc7 gives black a bigger advantage of almost 1 pawn so it is not going to like this position if it has an alternative to get equality. Uri
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.