Author: Uri Blass
Date: 23:09:05 11/17/03
Go up one level in this thread
On November 17, 2003 at 21:17:37, Kevin Heider wrote: >After watching Fritz not play either f5 or h5 in game 3, I have to say that >Deeper Blue (DB2) was the best chess computer EVER. Even with a better >evaluation algorithm; Fritz at 4m nods/sec can NOT be as strong as DB2 at 200m >nods/sec. > >DB2 simply knew when it had to start mixing it up. You cannot stop play and claim to be king. 1)Fritz is commercial so by definition kasparov can prepare opening that it does not know how to play. DB2 was never commercial and kasparov did not have a copy of something similiar to it. 2)Fritz chose a opening that is marked in the opening book as never played. Maybe they wanted to surprise kasparov and choosing the normal opening book could also lead to prepared win of kasparov. Deeper blue book makers did not need to care about that problem. lead to a win that was prepared. > >I took the position from Move 37 of Game #2 of Deeper Blue vs Kasparov. This is >the position where Gary claimed that someone (Karpov?) was feeding DB some of >it's moves. Gary could not believe it when DB played Be4! blocking the black >e-pawn from advancing. > >I had Rebel XP start an infinite search on my DURON 1100MHZ and it took almost >20 hours to find 37 Be4! (The move Gary claims a computer would never come up >with) I also did some analysis of game 2 and every commercial program that I tried expect Qe3 in its pv (in most cases after at most some minutes) and deeper blue had no idea about that move and did not expect kasparov to make it. There are even programs that find Kh1 instead of deeper blue's Kf1 at tournament time control. I remember that I did some comparison between Fritz and deeper blue in few tactical positions and it was obious based on the speed of my computer that Fritz with fast hardware is faster than deeper blue(most positions are not position when there are tactical line so it is hard to compare but I remember positions when there was a fail low of deeper blue in game 1 (deeper blue played a stupid move because of a bug but other program fail low for the same reason(Ke3) after some minutes and I think that comparing the times of fail low is relevant. > >Granted different programs do different searches/evaluations, but if Be4 is as >positionally strong as Gary claims then a lot of programs should like it after a >very deep search. > >If I assume: > >1. Rebel / Fritz / Junior / Shredder are of roughly equal strength > >2. DB took 10 minutes to find Be4 > >then I can conclude that X3D Fritz needs a better function or needs to run >(20hr*60m/10m) 120x faster to play at DB2 strength. > >Since I only own Rebel: Has anyone recently checked this famous position to see >if Fritz, Junior, or Shredder like the position after a deep think? > >Rebel XP's lines: >01:58:56 19.00 Qb6 Rxa2 Rxa2 >08:14:57 20.00 Qb6 Rxa2 >19:26:24 20.05 Be4 Rcb8 >32:42:43 21.00 Be4 Rcb8 Ra6 Qc8 Qa2 > >You will also notice on the right side of my screen grab that Rebel's 4th >favorite variation shows that if 37. Ra6 e4! Qe3 > >You can see a screen grab of the analysis at: >http://home.comcast.net/~kpheider/DB2BE4.GIF > >-- Kevin Heider I remember that there were discussions about it in CCC and it was not clear if Be4 is better. 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.