Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Strength, computers, GM's, ...

Author: Sune Larsson

Date: 05:23:25 07/17/01

Go up one level in this thread


On July 17, 2001 at 07:47:56, José Carlos wrote:

>  Strength is not an absolute magnitude. We can say player A is stronger than
>player B, and at most, we can figure how much stronger is A over B. But we don't
>have a mathematical definition of strength that fits every subjective definition
>people have. For example, if we ask GM’s (experts on the matter) who was
>stronger in his best time, Botwinnik or Capablanca, we’ll hear different answers
>with different reasons supporting them.
>  If we try to define strength difference in terms of results, we have the
>problem of the number of games. To know with enough certainity degree the
>relative strength of a pool of players, we need thousands of games between them,
>that we don’t have. In the best case, we’d get the _relative_ strength of _that
>pool_ of players in _that exact moment of time_. We know human players vary in
>strength from one day to the next. Additionally, we couldn’t compare any of
>those players with any other outside the pool. The definition of strength as
>results implies, IMO, that it’s nonsense to compare players that have not played
>each other, because the results against third players are in _different
>contexts_, so it means very little. That leads to the fact that we can’t compare
>players of different epochs this way.
>  So, it’s time to have a look at human tournaments and matches. In the past (I
>prefer not to consider the present situation) the world champion was decided by
>a very short number of games. First in zone tournaments, candidates and then
>against the previous champion. He didn’t play _all players in the world_ neither
>enough _statistically significant_ number of games. And we had no problem
>considering that player the world champion.
>All of this I’m speaking of, is just to try to show that it’s impossible by
>definition to speak of _absolute strength_, and that we humans have not bothered
>too much until now in the correctness of the practical relative strength.
>  Personally, I don’t find any interesting this (useless?) search for strength
>measurement, neither for computers nor for humans. I stick to the old fashion
>tournaments where everything was possible; where a weak player beating a strong
>player was a ‘big surprise’ rather than a ‘statistical event’; where the
>champion was the champion, no matter his ELO.
>  I’m afraid I should’ve been born in other time... in the past. :)
>
>  José C.

 Very interesting. As for computer strength vs humans, the easiest way is
 to look at the results. For me, Tiger's performance in the Magistral was
 just like if a 2600+ rated GM, in good form, would have played this tournament.
 Actually, if someone had showed me these games, without me knowing about the
 tournament, I would have believed that "the 2600+ GM" showed his talent and
 class. Computer's play a different sort of chess then humans. And this is
 more clearly to be seen in comp-comp games, IMO. Now we talk about "bridges"
 between comps and humans, antihuman mode and playing styles. This development
 is interesting.

 Sune




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.