Author: Peter Fendrich
Date: 13:05:57 06/04/99
Go up one level in this thread
On June 04, 1999 at 04:23:10, blass uri wrote: >suppose that there is a program with the name drawer. > >This program can always find the best move but it has only one bug. >If the opponent offers a draw then drawer except. > >Suppose humans know about this bug. >What is the rating of drawer against humans? > >If drawer play against kasparov it will do draws in all the games and have a >very high rating. > >If drawer plays against 1400 players then it will have rating of 1400. > >If drawer plays against computers then it may win program A 20:0 and do a draw >against a weaker program B because B offers the opponent a draw when the >opponent has an advantage and A does not do it because the programmer think that >it is insulting to offer the opponent a draw when the opponent has an advantage. > >The customer may think after seeing the results of A against drawer and B >against drawer(without the games) that B is better than A when it is clear that >A is better than B against humans and also against computers without the bug of >drawer. > >Uri Rating is based on randomness. The outcome of games are randomly spread out, surrounding the expected result. Whenever we have a situation without this random behaviour the rating formulas are not applicable. In fact it is quite easy to compute and predict the outcome of games between humans and this machine so rating is not even needed here. It is however not a very common case I presume... Another examples from the good old days when computers always played the same move in the same position and didn't vary its openings. Then it was easy to copy some winning line and play it over and over again, winning all the time. This is also a case when the rating formulas cant be used. //Peter
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.