Computer Chess Club Archives


Search

Terms

Messages

Subject: Rybka nodecounts; also, Comments about Arpad Elo's famous book

Author: Stephen A. Boak

Date: 20:45:54 02/16/06

Go up one level in this thread


On February 16, 2006 at 13:22:45, Zappa wrote:

>On February 16, 2006 at 01:26:41, Sune Larsson wrote:
>
>>Surely there must be some difference between "bad taste" and questioning a
>>persons *honesty*?
>>
>>Then again - who cares about nodes? The playing strength is what counts.
>>
>>/S
>
>I've always said Rybka is a great program, and ultimately what his program
>chooses to report in terms of nps/pv/etc/ is his business.  I'm just curious.
>
>anthony

Some comments--

Assuming your assumption is true:

1. Why not use a Rybkamark, like there is a Fritzmark--simply as a comparison
figure used for platform to platform, operating system to operating system,
compilation to compilation comparison.  Why not make it unique to your own
program, to avoid ignorant or unfair comparison to another program that does
things differently.

2. Why do some ELO lists of comp-comp results express strong program ELOs in
terms of hundreds, i.e. they don't assign ELO figures in the 2400-2900 range?
Are those ELO lists any less valid for the purpose they are provided (to show
relative strengths of the listed engines)?  In fact, they may be more valid by
*not* suggesting that engine ELOs from comp-comp play are somehow equivalent or
perhaps *identical* to ELOs of humans in the FIDE (or other) ELO scale.  And yet
they still properly show relative ELOs among members of the engine rating pool.

By the way, CCC readers, several years ago I had loaned out without return, or
otherwise lost my original book by Arpad Elo, "The Rating of Chess Players, Past
and Present", which I had owned for many, many years.

Just received another copy this week, per used book purchase online!

Great to be reading it again.  Fantastic description of the ELO system by Elo
himself--including the mathematical underpinnings.

This book is written very well and is very easy to understand, IMO.  Very
informative.

I recommend this book to anyone with a need or desire to learn more about the
Elo system (pros & cons) and statistical theory it is based on.

Elo discusses the strengths & weaknesses, as well as limitations of ratings
systems, including his own Elo system.

If you don't like math, or are not good at it, you will still be able to learn a
lot about the Elo system from this book.

If you are good with math & statistics, you will love this book and learn even
more.

Many CCC posters should read this book before posting their typically inaccurate
comments about statistics, or about Elo rating systems and their applicabilities
(or inapplicabilities).

--Steve



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.