Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Maximum ELO

Author: KarinsDad

Date: 13:10:36 06/09/99

Go up one level in this thread


On June 09, 1999 at 15:32:33, Christoph Fieberg wrote:

[snip]
>
>My guess is that in less than 50 years the best human player is at Elo 3000
>(because humans will learn from computer analysis) but will lose 90-10 against
>the best computer then.
>
>HOWEVER, I also believe that this computer then is still FAR, FAR away from
>playing perfect chess. Some hundreds years (or even some thousands years) from
>now on there MIGHT be the chance that chess is fully analysed.
>
>Perhaps it will be a surprise to discover that the ground position is won for
>White.  The computers who steadily increased their strength and knowledge over
>the centuries increased also their Elo rating. When the perfect machine appear
>the level might be at 7000. My guess. What is yours?
>
>Best regards,
>Christoph


I tend to agree with you with the exception that I think programs will win more
than 90% of the games from the best players in 50 years (probably closer to
100%). 50 years ago, computers didn't really exist (for all intents and purposes
other than as giant abacuses). Today, they are challenging GMs.

My guess on maximum computer elo, about 3500. The reason is that they may change
the equation (e.g. the USCF is changing their rules for provisional ratings this
year) such that it is impossible to improve. And, they will probably have a
different rating system for computers than they do for humans since I feel that
computers will be SO MUCH better that people cannot even compete with them, but
just use them to analyze games. It might get to the level of practically
memorizing variations out to 40 moves or so (yikes!!!), or you are at a
disadvantage.

People may get close to 3000 (Fischer once had a performance rating over 3000
and Kasparov had a performance rating over 3000 part way through a tournament
earlier this year), but I think that maybe 2920 or so is about the peak for
humans given the current rules. The reason I doubt it will get too high is that
all human players have the advantage of computers, not just the world champions.
So, the world champions will still draw (if not lose) enough that their ratings
will not soar TOO high. It will be the most talented people that are the
hungriest for it, just like always, that can make it to that level, but that
level will still be extremely difficult to achieve, even for a prodigy.

I also predict that if a program ever does solve the game (an extremely unlikely
occurance even assuming the most theoretical of computation possibilities), that
white has a forced win. The reason: White has the first move and Black will
always be playing catchup (in a perfect game).

KarinsDad :)



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.