Author: KarinsDad
Date: 09:48:37 07/02/99
Go up one level in this thread
On July 02, 1999 at 12:15:04, Paul Richards wrote: >Starting a new thread based on: > >On July 02, 1999 at 11:39:51, KarinsDad wrote: > >>It was a nice run by Topalov. And, of course he had help from Svidler who beat >>Fritz in round 12 and Morosewic who drew Fritz in round 13. With the exception >>of Polgar who seems to have an extremely difficult time even drawing against >>Fritz, it looks like the superGMs started getting Fritz' number (albeit a little >>too late). Svidler beat Fritz by allowing it as black to effectively play white >>with 1. a3 and surprisingly, playing an open game where he was able to push >>pawns and force Fritz to trade often (including Fritz trading down on the >>exchange of a rook for a knight). A very interesting game. > >This sort of thing begs the question of how strong the computer programs really >are against humans. A lot of success against the programs seems to depend not >so much on chess play but on use of special knowledge of the opponent, >specifically the knowledge that the opponent is a computer and therefore has >predictable strengths and weaknesses. This is the advantage of possessing >intelligence, but does it really show who plays better chess, or more accurately >does it really gauge how well the programs play? > >If a GM thought he were playing against a human I believe the program results >would be significantly higher than what they are now, even at long time >controls. It might be a bit underhanded to put together an event to test this, >so I don't think it will happen, but it would be interesting. Who was it that said that there are lies, damn lies, and statistics? There some inherent problems of playing programs against humans that most people either do not know about, or tend to ignore. Programs do not play erratically due to illness, fatigue, or psychological pressures. People do. People can adapt to their opponents and surroundings. Programs cannot (with the exception of some minor opponent adaptations). So, people come up with the idea of "tests" such as yours and make the implication that programs play better chess. However, take the following into consideration. Programs cannot ALWAYS play good chess without an operator. Someone has to be there to hold their hand in case they fall down. I could make the counter proposal of a long tournament test where operators are not allowed to assist programs. With this tournament, I am putting a limitation on the programs similar to your proposed test where you put a limitation on the human advantage of knowing your opponent. So, if there is a power failure and the program does not come back up with no assistance and pick up where it left off, too bad. It loses on time. If the program is connected via the Internet and there is a problem with the ISP, too bad. It loses on time. If there is an earthquake and the program is not smart enough to move out of the way of falling debris, too bad. This would give humans an edge over their electronic opponents. Now, of course, this is mostly tongue in cheek. But the main idea is that programs and people have different strengths and weaknesses when it comes to playing chess. By minimizing the strength of one side, you do not measure chess playing ability. Rather, you skew the data before you even start. KarinsDad :) PS. In the Svidler/Fritz game, Svidler also opened the game up which is not normally done against a computer, but Svidler saw he had the advantage. So, there are no hard and fast rules about anti-computer strategy just like there are no hard and fast rules in other areas of chess playing. Just guidelines. You still have to play the game.
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.