Author: blass uri
Date: 01:56:12 05/23/99
Go up one level in this thread
On May 23, 1999 at 01:52:34, Paul Richards wrote: >On May 22, 1999 at 21:19:11, Robert Hyatt wrote: > >>On May 22, 1999 at 18:15:57, Chuck wrote: >> >>>For White's 26th move Fritz 5 scores Bg2 -0.09 at 14 ply. >> >> >> >> >>This is a rare case of Fritz blowing a tactical position. White is totally >>busted here, and it doesn't matter what he plays. Let it continue to search >>and I'll bet it fails low at 15 or 16... > >No such luck. Just got to 16 ply, and it actually gives a plus score >for white after 26. Bg2. It gives ...Qc2 27. Qf1 Bxg2 28. Kxg2 Qe4+ >29. Kg1 Qxd4 30. Qe2 f4 31. Rxb7 and gives it a score of 0.12 pawns >in white's favor. As soon as 26...Qc2 is made, it goes to -0.22 >at 14 ply. I never understand why Fritz does this sort of flip- >flop in eval. You follow the line it gives and after you make the >supposed best move, the eval changes based on something it should have >seen earlier. In other words it searches to 16 ply, then you make >the move it suggests, and suddenly by 11 or 12 ply it gives a >decidedly different eval. Maybe this is a selective search thing? No The reason is that Fritz is a stupid root processor and the evaluation function depends in the oiginal position. It may do stupid mistakes because of it. It lost one of the games in the last WMCCC because it traded queens instead of playing for a draw. Only after it traded the queens it suddenly understood that the opponent has an advantage and not because of searching more plies forward. > >Anyway, lame human that I am, my idea after 26. Bg2 is ...Qc2 27. Qf1 >and then Qxb2, threatening to pin the queen with the rook. Lo and >behold once I give Fritz this idea, it agrees that black is winning >(although I don't know if I trust Fritz' eval anymore) ;) >So for clarity the line would be 26. Bg2 Qc2 27. Qf1 Qxb2 28. Bxh3 >Ra1 29. Rxf5 Rxf1+ 30. Bxf1 Kg8, with an eval of -1.75 at 13 ply. >Don't know if this is the best line, but I'm suddenly starting to >feel smarter than Fritz. ;)) 27.Qf1 is wrong and 27.Qe1 is the right move. 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.