Author: Dann Corbit
Date: 12:13:42 04/20/01
Go up one level in this thread
On April 20, 2001 at 04:37:46, CLiebert wrote: >On April 20, 2001 at 00:26:01, Adolfo Bormida wrote: >>Mayer-Kahlem wrote >>>The final version of the program playing against Kramnik must be made >>>available to Kramnik end of July already. This makes an match with equal chances >>>impossible ace it will be very easy for Mr. Kramnik to prepares and win to match >>>against any program under this circumstances. I would even say that it is >>>possible for human players of my strength to win to match against any program >>>with this conditions. > >As far as I know this is not right (without warranty). Beginning end of juli >Kramnik should always get the latest version of his challenger, but the >programmer could work further on his engine all the time! >Also Kramnik will not get the tournament-book and he also will not use the >8-CPU-System for his preparation. And I guess it would be quite different to >play with Fritz, Junior or Shredder (?) on a single/dual system or on >8-CPU-system. If he plays slowly enough, the differences will be microscopic. Since chess is exponential, at some point an additional ply will take so long that it simply won't happen. Why are computers without question the world's best blitzers, very very good at 40/2 and only average at correspondence chess? Because chess is exponential for computer algorithms. Let's suppose that he has a machine with two CPU's at 1 GHz instead of 8. The opposing machine will be four times faster. If you have a branching factor of 4, then it would give you one more ply. Two more plies tops. So you play the game 4 times more slowly (40/8hrs instead of 40/2) and you will get the exact same decisions if you want to wait that long. In any case, what is being sought is a strategy that will cause problems. If it causes problems at 40/2 on a slower machine, odds are good it will cause problems on a faster one. Extra horsepower adds less and less at longer time controls. Look at the mind-numbing computation power of Deep Blue. Why wasn't the ELO 30000 since it computed a billion raw nodes and 200 Million per second for actual computations? Because of the exponential nature of chess. I think Kasparov and Kramnik are about the same level. If Kasparov had 3 months to prepare with a Deep Blue prototype machine, how do you think he would have fared against Deep Blue? A little better, maybe? Perhaps a lot better. And then again, maybe the darn thing would have scared the pants off him and really psyched him out. ;-) Anyway, preparing against a computer opponent is a big help. Even with some randomization, their play is pretty deterministic. I don't think "wacky openings" will help though. That's one thing Kasparov tried (and it didn't help). It may have been a good thing against computers of the past, but I don't think it helps a lot anymore. You will find that (for the most part) when you play a computer at 40/2, if you remove its opening book, as you play it will develop along the lines of one of the standard openings anyway. When your opponent is looking ahead a full seven moves, they are going to develop solid formations.
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.