Computer Chess Club Archives


Search

Terms

Messages

Subject: Improvements of Chess Software

Author: Maurizio De Leo

Date: 18:32:26 08/11/03


I was interested in the development of the computer software in the last years,
putting aside the influence of the hardware improvements; this especially after
some result pubblished here were old programs (Nimzo, Genius;MChess Pro) battle
out well with new star.

Also I wanted to know how much Fritz 8 (for example) is better than Fritz 2 or 3
: is the improvement continuos ?

Unfortunately the most reliable rating list (SSDF) doesn't offer this
possibility. So I calculated a "normalized" rating list from the SSDF, with a
method that seems sound :

1) Only Pc programs,from the 386 onward. Every rating is normalized as if the
program had been tested on an Athlon 1200.
2) For every engine version the minimum possible number of conversions has been
made.

3) The rating gain between 2 computers is extimated as the average rating gain
of other engine versions that were tested on both harware
4) If no other engine versions were tested on both computers, the average rating
gain of all programs tested on both computers is used.
5) For every conversion it's added an uncertainity which equals the double of
the rating gain's standard deviation for those two computers, or it's extimated
in 25 elo if the data isn't enough.
So, if anyone is interested in the result e-mail me and I will send you the
Excel sheet, comprensive of an improvement graph.

Maurizio



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.