Author: Thorsten Czub
Date: 00:23:09 08/02/98
Go up one level in this thread
On August 01, 1998 at 10:42:10, Don Dailey wrote: >Thorsten, > >I think there is a factor in chess programs that are hard to determine. >It may be more important to have an "integrated style". For instance >your program may not be very good in most positions but perhaps it has >an unusual talent to avoid these positions it hates? I believe it's >hard to judge a programs positional play by a single number, but instead >have to look at the whole, does it understand the positions that it >actually gets in games? Does it hide it's own weakness well and does >it force it's own strengths on the opponent? In Paris fritz was UNABLE to understand the positions it played. This fits very much to my impression of the program. Most of the games Fritz wins, the opponent is outsearched. If you compute 2 plys deeper than your opponent, it looks as if you play positionally better. But in fact, you don't. In the draw/loss games of fritz you can study this draw-bug problem. Markus Gille (DarkThought) and I studied this behaviour in a tournament game here at my home. The behaviour came not only in ONE move, the strange moves, main-lines and weak moves were a few moves. Fritz struggled. And the opponent was able to use the draw-behaviour to make a draw (when the opponent is -) or to win (when the opponent is +). Such a behaviour was typical in paris. I have spoken with Dirk Fr. and Moritz B. about this behaviour, they all know what I mean because they have seen it on their machines too. IMO fritz sees silly repetition lines deep in the tree (because it extends deep check check check lines), because whenever the program has this behaviour, it does not come much deep, only 9 plies, and behaves very slowly. >When I used to play in tournaments I noticed that each player has his >own style. Style is a quality feature ! So I know style will affect playing strength and behaviour. > I often found myself getting the same kinds of games with >each opponent. Some of these games it was quite clear the opponents >knew how to play this kind of game, and yet their whole chess style >seemed integrated around specializing in this kind of game. I would >see them playing other players and they always were getting the same >types of positions. > >Have you ever played non-consultation doubles chess? It's a lot of >fun, you play with a partner and you take turns moving without >consulting each other. You will find the actual quality of >the games to be unusually low, each player is persuing a different >logic about how to play the game. Two 2200 players will probably >play 2000 chess doing this. It would not surprise me if two >computers playing in this fashion play much weaker than the >weakest individual computer. Good example. Fritz plays this way ! :-))) >I really believe it's more about doing what you do best, and >doing this well. Most good chess players are aware of very >ideosyncratic players who get excellent results. I think this >is why, they have specialized their game to a remarkable level. > >- Don
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.