Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Can someone make a program to ultimate perfection at 20 hrs. per move?

Author: Sune Fischer

Date: 15:37:47 03/19/02

Go up one level in this thread


On March 19, 2002 at 17:53:25, Daniel Clausen wrote:

>Hi
>
>On March 19, 2002 at 12:52:13, Sune Fischer wrote:
>
>[snip]
>
>>The elo system can handle any result you throw at it AFAIK.
>
>Well yes, but it's designed to determine relative strengths between players. And
>they should be somewhat close rating-wise. The exact formulas vary a bit from
>country to country (well ok, FIDE is more standardized :p) but basically you
>don't get any ELO points if you're - say - 750 points better than the rest in
>your pool of chess players. Therefore the best player couldn't get more points
>than 2850+750 ELO, even when beating Kasparov each time.

Well, one could just use the real gaussian and count frational points, even they
will add up :)

>>What needs to be defined is "perfect play". For instance will the perfect
>>player just pick the moves random from those that lead to the draw, like an
>>engine with no evaluation other than the score for win/draw/lose, or will he
>>always choose the longest most complicated game possible (like swindle mode).
>
>The perfect player doesn't need swindle mode, he/she/it just beats you and me
>(and Kasparov) no matter what. :p According to my definition a perfect player
>knows that the opening position is mate in X, mated in X or draw. Stuff like
>swindling mode are beyond it.

You forget that chess is a limited game, even by pure statistics the perfect
player will not win _every_ game.
I think it will matter a great deal if the perfect player understands, that
prolonging the game or complicating it by using agressive gambits will be in its
favor.

-S.

>Sargon



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.