Author: Robert Hyatt
Date: 09:28:00 11/25/97
Go up one level in this thread
On November 25, 1997 at 04:00:43, Ernst A. Heinz wrote: >On November 25, 1997 at 02:04:53, Howard Exner wrote: > >>Cilkchess is doing nicely on the scoresheet at the Dutch Championship >>but it looks to me like the chess gods are smiling on it just a bit. >> >>8/p3pp1p/6p1/4P3/2K2P2/6P1/1k5P/8 b - - id"Cilkchess 2.0 - Nimzo 98"; >> >>Here Nimzo98 played h7-h6 but doesn't the move Kb2-c2 put the nail >>in the coffin? > >Yes, it does -- most chess programs choose the winning Kc2 in a couple >of seconds (DarkThought eval = +1.3 after 12 sec). > oversight or evaluation bug or selective pruning error. Kc2=+.69 until ply=11, then +1.00 through ply=13, hits +1.5 and fails high on that at ply=14. >> >>6k1/pp3p1p/2p1p3/P1P2p2/8/1n2P3/1bR2PPP/6K1 w - - id"Shannon - >>Cilkchess"; >> >>Here too the Cilkchess team must have been relieved when Shannon >>played a5-a6. The straight forward Rxb2 would lead to a win and a >>huge upset given the Cilkchess Hardware. This 2nd position would >>perhaps be more difficult for computers to eval since the material >>is balanced. But the Knight is no match for the rook in this ending >>especially with that weak isolated h pawn. > >I do not know if Rxb2 is really won but it obviously offers the best >winning chances. DarkThought sticks to Rxb2 right from the beginning. > >=Ernst= Crafty also would only consider rb2 here, and I let it run for many minutes. But I'm not convinced this is winning at all. It seems better than anything else, but the difference between a knight and rook is often less when a computer has the knight, because it can calculate very accurately, while a human can make a mistake. And with a knight, a single tempo can be critical. the eval here was steady at around -.2 through 15 plies (on my notebook) before I stopped it. However, none of this is a surprise. You will find the same sort of good/bad luck in games at Paris, Jakarta, Aegon events, etc. There's a lot of "luck" involved, although better programs generally have better "luck".. :)
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.