Author: Tony Nichols
Date: 18:52:28 10/19/04
Go up one level in this thread
On October 19, 2004 at 10:00:20, Tord Romstad wrote: >On October 19, 2004 at 02:56:31, Tony Nichols wrote: >> >>When a GM plays against a computer in the opening he's actually playing against >>other GMs. You could a chess program think for a month and it's never going to >>play the first ten moves of the Najdorf! > >I've never quite understood this argument. If you somehow removed a GMs >memory of opening theory and allowed him to think for a month, he will also >not play the first ten moves of the Najdorf. GMs play the Najdorf because >they know the theory built by the work of hundreds or thousands of players over >several decades (or, for several other openings, more than a century). When >human players are allowed to stand on the shoulders of giants, why is it not >fair to allow computer players the same? > With this logic no one would have ever invented the Najdorf. >Some people argue that it is unfair because the computer has perfect and >practically unlimited memory, and can remember all lines ever played in the >Najdorf. But on the other hand, human players have other advantages. A GM >playing the Najdorf has an enormous amount of knowledge of the plans and >the tactical and positional motifs of the opening. He knows where to place >the pieces, which pawns to advance, and which pieces to exchange. He has >detailed knowledge about the typical endgames resulting from the opening. >The computer has none of this knowledge, and has to work everything out on >its own from the moment it leaves its opening book. > Humans also do this but they start on move 1. They can use ideas from other GM games but they have to understand every move. Chess engines would never play the openings they do without a book. So basically we have human players playing the opening for the computer. Then the engine can play the middle game but when the endgame comes humans again step in and the engine is not allowed to play the moves it would choose. This is fair? >It is true that when computers grow faster and stronger, it might be >interesting to play human-computer matches where the computer is handicapped >in some way. But it seems ridiculous to me to regard such matches as more >"fair" than the traditional format. Computer and human players simply have >very different strengths and weaknesses. Taking away some of the computer's >strengths while allowing the human to keep all of his does not make the game >more fair. > >Tord When we speak of human vs computer matches the term traditional hardly applies. We have simply been using the format for human vs human matches. As regards the opening book. I think it is the equivalent to letting GMs consult opening materials during play. The same goes for endgame tablebases. These things are excepted as part of the "chess program". I think the real question is about the strength of the engine. Anyone can make opening books, and most programs use the same Nalimov endgame tablebases, so there is no skill involved from the programmer for these. When we talk of program X beating GM Y sometimes it has very little to do with the strength of program X's engine.I think the best thing that can be said about chess engines is they don't blunder in the middlegame. I think the only logical way to view human vs computer matches is from the perspective of whether or not engines are getting stronger. In this regard opening books and endgame tablebases are detrimental to seeing the true value of the engine. I think we need to redefine what we consider fair for these matches and why we even have them. Regards Tony
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.