Author: Robert Hyatt
Date: 20:06:16 07/14/98
Go up one level in this thread
On July 14, 1998 at 18:46:53, Shaun Graham wrote:
>On July 14, 1998 at 17:28:23, Bruce Moreland wrote:
>
>>I think the argument that a computer is a GM as long as it can play from behind
>>a black curtain is awful.
>
>
>That is not the arguement. The arguement is simply that to test accurately how
>a computer program(Fritz) would perform against standard play in a
>tournament(swiss), for SCIENTIFICLY acceptable data you would need to remove the
>bias, and that bias requires that the opponent not know he/she is playing a
>computer. The Hypothesis, is that indeed a program such as fritz would perform
>extremely well.
you keep making that statement, but it is *dead wrong*. In a scientific
experiment, you don't weed out circumstances you don't like. Because then
it becomes a contrived experiment. IE you can travel at C in a perfect
vacuum? Very good. But where do you find that perfect vacuum? Not in
outer space? No vacuum there (plenty of atoms to run into, even if you go
to inter-universpace where you might only find one atom in a cubic meter.
But you can't take important considerations out, just because they are
inconvenient. Humans play each other effectively and alter their strategy
based on lots of things. A computer has to be able to handle this, or else
you get this:
fritz GM*
(* not a real GM, a GM if you don't allow the opponent to know it is playing
a computer, or else you don't allow the opponent to play anti-computer type
strategies against it).
Now, is that a GM, or what? I say it is *not* a GM. This is the typical lazy
out of a problem. Your opponent keeps beating you for playing the same opening
over and over? Just don't count the duplicate games. Don't do any work to stop
playing the same losing line over and over, just don't count them. Your
opponent smashes you by playing closed positions? Don't solve the problem by
making the program play the positions better, solve the problem by either not
letting the human play closed positions, or try to hide behind a curtain so your
opponent won't know that he's playing a computer and won't try such strategies.
To be a GM, you have to be a *complete* player. You have to play tactically
strong, positionally sound, understand open and closed positions, be able to
play decent endgames, and so forth. If one part is missing, it is a hole that
other GM's will jump in and stomp until it is dry.
>>
>>Even in a weekend Swiss, where you might not know everyone, you at least know
>>that they are human and what their approximate strength is (their rating).
>>
>That's when you are not provisionally rated, a program entering a rated
>tournament for the first time would be provisional. We are not talking about
>making his opponent feel happy, we are not even advocating the experiment take
>place. We are simply discussing what it would take to demonstrate the strength
>of a program against standard play. If you are interested in how a program
>plays competitively, well that's fine, A program such as fritz can not be a GM
>competitively for the reason, that more can be known about it than what can be
>known about regular opponents, and it has not the ability to be flexible and
>deal with "anti-computer chess". However if the question is "When i buy a
>program to better develop my regular chess(as opposed to anti-computer chess)
>will that program perform like a grandmaster against my standard chess?" Thus
>the point is to find out is Fritz GM strength(2400-2500+ELO) against standard
>chess play.
>>bruce
I agree, but you are *not* describing "standard chess play" by any definition.
You are describing a "contrived circumstance"...
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.