Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Inflationary Effects?

Author: Uri Blass

Date: 15:07:50 07/09/03

Go up one level in this thread


On July 09, 2003 at 17:37:46, Russell Reagan wrote:

>On July 09, 2003 at 16:07:03, Jeroen van Dorp wrote:
>
>>An Elo rating tells you the difference in strenght between two opponents. Have
>>you any clue why the rating difference between shredder and the next on the list
>>is "not meaningful"?
>
>In addition to what others have said, someone posted on r.g.c.c that Shredder
>7.04 has also had an easier go so far.

I do not agree.
Weaker opponents does not mean that it is easier to earn rating.

 Shredder's opponent rating is the lowest
>of the top 21 engines tested. It hasn't been tested against Deep Fritz 7 yet,
>which the Shredder 7.0 had a very difficult time with.
>
>As Dann reminded me yesterday, there are error margins, which show that Shredder
>7.04 isn't statistically better, yet. So this tells us what we've known for a
>long time. The top engines are all approximately equal in playing strength.
>
>Shredder 7.04 2810 - 34 = 2776
>
>Shredder 7    2770 + 27 = 2797
>Fritz 8       2762 + 26 = 2788
>Deep Fritz 7  2761 + 28 = 2789

I think that it is wrong to add the error margins because the error margin for
the difference between shredder7.04 and Fritz is smaller then the sum of the
error margin of both programs.

I think that shredder7.04 is probably the best and if you calculate an error
margin for the difference between shredder7.04 and Fritz8 than 0 is not going to
be included with confidence of 95%.

An approximation can be done by dividing the sum of the error margin by sqrt(2)

60/sqrt(2)=60/1.41..<43
2810-2762=48>43

Uri



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.