Author: Ed Schröder
Date: 23:13:01 11/15/98
Go up one level in this thread
>>Hi Amir, >>I remember in the past we did some analysis with both our programs. I >>believe the outcome was: >>In one of the main lines Deep Blue sacrificed 2-3 pawns for a dead wrong >>king attack. We came to the conclusion that Deep Blue must have a very >>speculative king safety algorithm or the main line was a bug in the king >>safety algorithm as in the main line there was no compensation at all to >>justify the computer score. >>Maybe you can correct me if there are errors in the above as this is >>what I remember from our discussion. > >>About move 36. axb5 >>r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - id DEEP BLUE - >>Kasparov,G; bm a4b5; >>I have run 36. Qb6 overnight. After 16 plies Rebel10 says: +1.29 >>I will run 36. axb5 tonight. >>I understood this is the position you talked about? >Yes, you got it right. >The PV for Deep-Blue on all iterations except the last starts 36.Qb6 Qe7 >37.axb5 Rab8 38.Qxa6 e4 39.Bxe4 Qe5. In the last iteration, there's no PV. Thanks for posting the main variation. I clearly remember the end position again. Based on this impressive main variation I can come to no other conclusion other then that Deep Blue must have a speculative king safety. The other explanation, a bug, sounds not fair to the Deep Blue team. Based on this I would say that cheating did not took place at all. I mean if a program can come up with such (for a computer spectacular) main variations then it's also very logical to assume that it may change its mind later as IMO speculative programs tend to change the "best move sofar" all the time this due to the high bonus / penalties they give. So it's very likely that after going another ply deeper the black score for 39..Qe5 even more increased and therefore 36.Qb6 fail low. Then the program changed to the "safe" 36.axb5 instead. So I would say no cheat as suggested by Kasparov. - Ed - >This is the variation that Kasparov and others talked about. >I disagree with your description of the attack as "dead wrong". You give here >the "official" computer chess opinion on it. Kasparov and others GM's would >maybe play it without hesitation, on an "only chance" basis, and would call it >speculative rather than wrong. White is better, but black has a real game >going. >It is exactly this difference in the computer and human outlook that made >Kasparov doubt that the computer played axb5 on its own. >Amir
This page took 0.01 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.