Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The Nunn Book is not necessary to test personalities

Author: blass uri

Date: 06:24:23 06/24/00

Go up one level in this thread


On June 24, 2000 at 09:13:46, blass uri wrote:

>On June 24, 2000 at 07:48:59, David Dahlem wrote:
>
>>
>>On June 23, 2000 at 23:41:15, Jorge Pichard wrote:
>>
>>>On June 23, 2000 at 23:37:31, Jorge Pichard wrote:
>>>
>>>>On June 23, 2000 at 23:32:16, Jorge Pichard wrote:
>>>>
>>>>>On June 23, 2000 at 13:28:03, David Dahlem wrote:
>>>>>
>>>>>>
>>>>>>On June 23, 2000 at 11:18:33, Daniel Chancey wrote:
>>>>>>
>>>>>>>Your book, ~nunn.obk has only 403 book lines.  This is unacceptable to be used
>>>>>>>in tournaments or computer vs comptuer challenges.
>>>>>>>
>>>>>>>Castle2000
>>>>>>
>>>>>>The Nunn book was created for testing personality settings only, not for
>>>>>>strength. In my opionion, it gives more consistent results since neither side
>>>>>>gets an advantage from the opening book.
>>>>>>Dave
>>>>>
>>>>>The easier way to avoid an advantage without going thru the problems of >creating a specific book like the Nunn book, is to set an amount of games that >you are willing to test the personalities, for instance if your goal is to >only test it for 30 games, then let one personality plays 15 games with the >white pieces,then switch the personalities side of the board and replay those >15 previous opening, allowing both personalities to play 15 diferrent openings >with both sides of the board.
>>>
>>>PS: I forgot to mentioned that the first 15 games you simply allow the program
>>>to select whatever opening it desire to play and of course eliminate a repeated
>>>opening in the first 15 games.
>>>>>
>>>>>Pichard.
>>Yes, that is one good way to run personality tests. Using an opening book with
>>only equal lines is another. The ideal way to test is without using an opening
>>book at all.
>>Dave
>
>This is the worst way to test because if you use deterministic programs you will
>get the same games again and again.
>
>You may also get wrong results because some programmers did not include some
>knowledge about opening because opening book covers this knowledge most of the
>time.
>
>Uri
If you want to play without opening books you can give programs this illegal
position:

[D]RNBQKBNR/PPPPPPPP/8/8/8/8/pppppppp/rnbqkbnr w - - 0 1


I played some games in the past from this position but I do not think that the
result is important because it can tell you almost nothing about the ability of
programs to play chess.

I also think that playing without opening book tells you wrong information about
the ability of programs to play chess.

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.