Computer Chess Club Archives


Search

Terms

Messages

Subject: SSDF rating and Elo rating is not the same.

Author: emerson tan

Date: 05:41:46 07/06/03


SSDF rating and Elo rating is not the same.


If the computer, without humans adjusting its parameters and opening books,
plays thousands of games against several humans, the computer's rating will go
down because of its inability to learn and understand. It is limited to avoiding
opening lines it lose  and the specific positions if it has positional learning.
The Humans on the other hand can learn from mistakes and can easily improve
their play after each game against an opponent who cant learn. The human's will
keep on picking up ideas on where the computer is weak or has a disadvantge in
its play and the humans can formulate a plan to exploit them.

Maybe an IM can lower the rating of the computer by just playing for a draw. The
IM would just anticipate the attacks in advance ala petrosian and counter the
moves several moves in advance. If a human would just keep on playing the berlin
wall, it would be very hard for the computer to win. The computer may score a
lot of points at the begining, but after several hundred games the score of the
humans will improve.

The performance of computers in computers versus humans is usually good because
the GMs either is faced with new unknown opponen or the programmer can change
the program in between games. Not knowing how your opponent plays is proably
worth around more than 100 elo. Deep Blue have both the advantage. Its always a
new Rebel when it plays against a GM. Junior can be change in between games.
Boris Alterman himself said that allowing them to change the Junior in between
games is a really big factor.   In Kramnik vs Fritz, its more of Kramnik fault
than the good performance of Fritz. Anyway, the programmers still have to
intervene in the type of opening being played by Fritz.

Elie Agur mentioned before that the computer, being preprogramed, will ever fall
short of equalling man in one respect: the programs would not be paradoxical
enough to make its play unpredictable. The computer's predictability would
constantly enable human players to devise new methods to oppose it successfully.

Ofcourse the above only holds true until programmers can make a program that
really understand chess.



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.