program ELO ratings

[Steve Worboys]

Geschrieben von:/Posted by: Steve Worboys at 16 February 2000 05:24:21:
I would like to calculate ELO ratings for WinBoard engines. How should I go about this? Can someone advise me?
Sincerely, Steve Worboys

[Aaron]

Geschrieben von:/Posted by: Aaron at 16 February 2000 07:23:13:
Als Antwort auf:/As an answer to: program ELO ratings geschrieben von:/posted by: Steve Worboys at 16 February 2000 05:24:21:

I would like to calculate ELO ratings for WinBoard engines. How should I go >about this? Can someone advise me?
Sincerely, Steve Worboys

You probably need some software to do it for you.
But as i understand it such ratings tends to be meanless if you want to compare it to Human Elos..The value is just relative to the programs you test..So unless you have a engine that has a accurate ELO relative to Humans and use that at a base, I don't really see the point.
Is this correct? So while it is significant to say that Crafty is about 100 Elo stronger than say GNUChess but less more when saying Crafty is 2600, GNUchess 2500??

[Dann Corbit]

Geschrieben von:/Posted by: Dann Corbit at 16 February 2000 20:01:20:
Als Antwort auf:/As an answer to: Re: program ELO ratings geschrieben von:/posted by: Aaron at 16 February 2000 07:23:13:

I would like to calculate ELO ratings for WinBoard engines. How should I go >about this? Can someone advise me?
Sincerely, Steve Worboys

You probably need some software to do it for you.
But as i understand it such ratings tends to be meanless if you want to compare it to Human Elos..The value is just relative to the programs you test..So unless you have a engine that has a accurate ELO relative to Humans and use that at a base, I don't really see the point.
Is this correct? So while it is significant to say that Crafty is about 100 Elo stronger than say GNUChess but less more when saying Crafty is 2600, GNUchess 2500??

An ELO calculation requires a large number of games against a pool of known opponents. The larger the number of games, the more secure the ELO figure is. The absolute number of an ELO has practically no meaning. The value of an ELO number is in differences. For two opponents in a pool of players, if we know the ELO difference we can make probability guesses as to the outcome (over the long haul).
By cracky, I do believe you've got it. It's the difference that counts.
Here is an ELO calculation program (for USCF ELO), and tournament arranging programs:
ftp://38.168.214.175/pub/tournament_software/

My FTP site