Author: Mike S.
Date: 18:29:02 11/19/00
Go up one level in this thread
On November 19, 2000 at 14:08:06, Bob Durrett wrote: >(...) >We need chess-playing computer programs which are smart enough to come up with >opening moves which are as good as, or better than, those produced by the top >GMs, without resorting to stored opening books. [Especially true if those >books are merely collections of games produced by mere humans.] > >And we need chess-playing programs which never play dumb moves, or almost >never, i.e. dumb only 0.0001% of the time. > >Until that time comes, there is still a need for improvement in chess-playing >programs. Not to play against, but to use for analyses. But how strong a player does one have to be, to determine the difference between this future situation, and the one which we have already? For example, my impression is that Rebel - when it's out of book early - can calculate opening moves which appear absolutely classical to me. High level quality opening moves, as if they came right out of a book by Tarrasch, or by Nimzovitch when he talks about developement, tempo, etc. And of course the moves are tactical exact to 99.x %. Or, if the situation has only the least degree of sharpeness, let Nimzo calculate the continuation. It will most always find and play the very sharpest opening continuation you can think of. Given 3 minutes on a up-to-date pc, no GM can win a game against these programs in the opening I think, not even when the computer has only the ECO positions as a book available. Of course you can't blame a computer program for falling into a trap then and when, which took mankind 100 years to discover and develope. This can happen to a GM too. Books from game collections are a constant topic (mistakes...). But for chess programmers is important, that such large books will increase their chances in 99.x % of the games, and therefore they can afford that an "accident" may happen in 0.x % of the games. Chess connoisseurs would probably prefer "theory" books consisting only of moves, which have gone through an analytical filter of a human master and opening expert, or come from his analysis. I think that the small books from the early time of computer chess were made like this, but now it is usually a mixture of both concepts. I'm afraid we'll have to live with that, because for chess programmers, success in terms of rating list places and tournament results it is most important. They cannot sacrifice some of their percentage to the strange wishes of too demanding customers. Regards, M.Scheidl
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.