Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Feng-Hsiung Hsu's talk at Microsoft

Author: Uri Blass

Date: 15:41:36 10/08/02

Go up one level in this thread


On October 08, 2002 at 18:25:08, Uri Blass wrote:

>On October 08, 2002 at 17:19:39, Angrim wrote:
>
>>On October 08, 2002 at 16:28:39, Robert Hyatt wrote:
>>>
>>>OK.   I have a quad free and am playing two games at a time using two copies of
>>>xboard,
>>>one crafty has normal null-move (2/3) the other has none (sel=0/0).  I will
>>>report as I get
>>>results...
>>
>>I have noted before that the effects of a change tend to be greatly
>>amplified when tested with self-play.
>
>I do not believe in that theory.
>
>I remember cases when programs did not earn rating against previous version in
>the ssdf games.
>
>If I remember correctly than Hiarcs7(p200)-Hiarcs6(p200) 20-20 is one of the
>results of the ssdf inspite of the fact that hiarcs7 is clearly better based on
>results against other programs.
>
>Uri

I can add that x+1 against x can be misleading but to both directions and I see
no reason to assume that the real difference is smaller.

The best test is to test x against some players  and x+1 against the same
players.

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.