Author: Mark Ryan
Date: 15:30:19 05/17/00
Go up one level in this thread
On May 17, 2000 at 17:29:02, Jeroen van Dorp wrote: >On rgcc Magnus Carlsson posted the following situation. He let Fritz 6(a) do a >blunder check of an ICC game and got a strange result. >In this game: > >Sleeper1 (1291) - MagnusCarlsson (1373) [B50] >ICC 2 12 Internet Chess Club, 14.05.2000 >[Fritz 6] > >1.e4 c5 2.Nf3 d6 3.Bc4 e6 4.d3 a6 5.Nc3 g6 last book move 6.a3 Bg7 7.Bd2 Ne7 >8.b4 b5 9.Bb3 Qc7? [-1.38 Fritz 6: 9...Nbc6= -0.19/8 ] 10.Rb1 cxb4 11.Ne2 >bxa3 12.Qc1 Bb2 13.Qd1 a5 14.0-0 0-0-+ [-2.25 Fritz 6: 14...b4 15.c3 Nbc6 >16.d4 0-0 17.Bh6 Re8 18.c4-+ -1.63/8 ] 15.Qe1 b4 16.Ned4 Nbc6??-+ [-5.34 >Fritz 6: 16...Ba6-+ -1.97/7 ] 17.Bc1 Bxd4 18.Nxd4 Nxd4 19.Bc4 Bb7 20.c3 >Ndc6 21.cxb4 axb4 22.Bb5 Ba6 23.Bxc6 Nxc6 0-1 > >Fritz made an amazing analysis, when *only* black was blunderchecked, at move >16. Basically Fritz states black did a bad move (didn't) and suggest a move that >is even worse. He classifies 16.Nbc6 with ??, give a -5.34 evaluation, suggests >Ba6 with evaluation -2.00, which is of course a worse figure for black. > >Basically 17.Bc1 is the mistake, as the knight (see diagram and play Bc1) is now >attacked twice and defended once. > >If you let Fritz blundercheck *both* sides, it doesn't happen, and it gives a >right assesment. I reproduced Magnus's results, and got the same . > >Is anyone familiar with the things happening here? > >The blundercheck was performed at 5 secs/move > >Thx. > [snip] Hi: In the past I have blunderchecked 100 games using Fritz 5.32 and 60 games using Hiarcs 7.32 at 60 seconds per move; I always checked both sides. I have frequently found the situation where the blundercheck suggests that a move was bad because another move was better, but the evaluation indicates the opposite - - the suggested alternative move is worse by blundercheck's own evaluation. I always assumed that this was a flaw in the blundercheck program. Cheers, Mark
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.