Computer Chess Club Archives


Search

Terms

Messages

Subject: About computers and ELO and TPR and stuff.

Author: Chrille

Date: 13:39:39 07/10/00


Hello everybody, I've been watching the discussions here about Juniors
performance in the Dortmund tournament, and I just want to make an observation.

A players ELO rating is an indication of how well this player has previously
played against humans.

Now, I don't think anyone will argue that playing a computer is a totally
different thing from playing a human; and that thus a players ELO rating does
not directly indicate how great an achievement it is (computerchess-wise) to
defeat this player.

A win against Hasidovsky (2200 computer-drawer in the Israeli league) is more of
an achievement than a win against Huebner (2650 computer-rookie in Dortmund).
Probably.

So lets not all get nuts now that Junior starts getting performance ratings
around 2800. This does not directly mean that the software has been much
improved since the Thisandthat tournament last month where it got a TPR of 2400.
I believe it is rather an indication that a humans strength against computers is
not as highly correlated to his ELO rating as we have previously believed. I
think TPR's are rather bad indications of a computers performance, as well as
any other application of a human ELO rating or title for that matter.

I don't even really look at the numbers. Just look at the games instead and you
will see that Junior has played some (computerchess-wise) great chess, I really
admire some of the moves it makes. You will also get a very clear image of how
it plays chess, what has been improved, and what can be improved, which is
really interesting to me.

Good night,

- Christian Söderström
http://ds.onyx.nu/mint/



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.