Author: Vincent Diepeveen
Date: 11:33:07 06/20/01
Go up one level in this thread
On June 20, 2001 at 10:19:19, Robert Hyatt wrote: >On June 20, 2001 at 08:17:57, Joshua Lee wrote: > >>Hi, I was looking at game 6 from Kasparov's first match against Deep blue When i >>noticed some analysis that i had missed before, that of Kasparov not playing the >>more forcing win. >> >>Kasparov,G - Comp Deep Blue Game 6 1996 >> >>[D] 2rr2k1/pp1qnppp/2n1p3/3p4/1bPP3P/1P2RNP1/PB3P2/1BRQ2K1 w - - 0 1 >> >>I had been wondering if any program could point out Deep Blue's mistake which >>leaves the eval at +- is Kasparov's favor however this is more gradual and there >>is probably no hardware in the world that can see that deep who knows how deep >>that would be??? Then i got the idea if commercial software could find the >>quicker win that i would have a better idea of what they really need to trounce >>Deep Blue . In reality though, this is another story as after 27 and a half >>hours and a depth of 20/53 Fritz still prefers Kasparov's move 20.a3 but the >>eval is wrong when it suggests 21.c5 so i take it that computers still need alot >>more work. > > >Hsu reported that DB expected the bishop sac and had a score of 0.00... I saw >a lot of GM analysis that finally concluded that Kasparov's move was better, as >after his move in the game, he threatens the bishop sac which would then win. > >I don't recall the analysis in detail however, just that it was discussed a lot >while the game was in progress and then later at several chess sites around the >world... The bishop sac is completely won for white, but initially it is vague compensation, only after that a passer on the h-file of white promotes unstoppable to a queen. This line is hard to work out of mankind whereas a3 b4 c5 wins chanceless. So it's logical that kasparov didn't do any effort to checkout Bxh7, but just shuffled a bit and won the game chanceless. The win is real deep. Like 25 plies or so to get a big score for the passers. Nevertheless Bxh7 is completely winning and probably too deep for most programs to find, as the passer advance near the end of the line is another 10 plies extra above finding the bxh7 trick. On the other hand with a bit optimistic king safety you quickly find Bxh7. I remember several DIEP versions finding this shot within a few minutes on a pentiumpro200 back at start 1997. It completely depends upon king safety terms you use and bonus for passers you give. i'll give it a try right now, but all terms of diep are pretty low tuned, so i'm not so sure it soon will play it as just like kasparov diep likes a3 b4 c5 to win the game chanceless and positionally, so i need a score of above 0.80 or so to find Bxh7.
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.