Author: Dave Gomboc
Date: 03:33:39 05/19/99
Go up one level in this thread
On May 19, 1999 at 05:57:24, Francis Monkman wrote: > >On May 19, 1999 at 04:20:27, Dave Gomboc wrote: > >>I agree again. :) This is especially true early on in theory. When you get >>nearer to the end of lines though, it seems to me that the human is wrong more >>often than not. One often sees "with a big attack (clear advantage to Black)" >>even though Black is down a couple of pawns or something, let the computer grind >>on it, and it seems to like White's position just fine. You try all the attacks >>you can think of, it handles every one. So now what is the story? Is the GM >>wrong? Or did I just miss the right continuation? Hubris that it may be :), I >>often conclude that the GM made an error. >> > >In many cases I've found this to be true. However, you only have to play through >a selection of games by eg Tal, or Smyslov (pick your favorite) and I guarantee >you'll find examples a-plenty where the machine is just wrong, calculating on >'surface factors', and where a human player of the stature of Tal etc can 'strip >away' the surface (knowing that he'll be able to exchange favourably etc), the >computer is unable to do this. Also you'll notice that so often a move that >genuinely transforms the situation is not even registering in the machine's top, >say seven, choices. > >I really think it's well worth walking your favorite analysis-engine(s) through >as many >top human games (old and new), as possible. I'm sure your faith in 'human chess' >will be rapidly restored! > >Francis No, it was doing this that broke my faith in 'human analysis' in the first place. :/ The (e.g. Informant) analysis is often tactically busted. I believe that some annotators set up these traps on purpose. One good thing about computers being tactically strong is that this sort of this will become less popular in the future. You might be right with respect to "best of" collections. The quality of analysis in those is probably higher than in the run-of-the-mill Informant game. Dave
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.