Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Who is the better chess program author?

Author: Ren Wu

Date: 12:03:12 12/12/01

Go up one level in this thread


On December 12, 2001 at 14:36:07, Dann Corbit wrote:

>On December 12, 2001 at 13:51:37, Ren Wu wrote:
>
>>On December 12, 2001 at 00:33:36, Christophe Theron wrote:
>>
>>>The least important kind of experts from the list above are the chess experts.
>>>
>>>Yes, sorry.
>>>
>>>
>>>
>>>    Christophe
>>
>>But who is responsible for making opening book?
>>
>>Looks to me that auto generated book (or book made by programmer) is not as good
>>yet.
>>
>>A chess expert at side can be valuable for this.
>
>I believe at some point, computer generated books will be better than human
>books.  We are a long ways from there right now, but I think some simple
>breakthroughs can change it.
>
>I have 24 million chess positions which have long time control analysis.  And
>200 million+ with about 9 plies.  If this data is mini-maxed, and combined with
>database information about what people do by statistical inferences to their
>ELO, the number of those with high ELO choosing the move, the won/loss/draw
>ratio from careful subsets of the data, etc. -- then I think a computer will do
>better than a human.  After all, that's what the humans are doing.  And the
>computer will make fewer mistakes.  The only difficulty is programming the right
>variables into the mix.  If the queries are performed in ANSI SQL, then it
>should be very simple to test a very large range of possibilities.

Yes i agree with you on most of the points.

There are some problems too,

1. It is hard to take the engine's strength/weakness into account when build the
book. some positions human can do great, but not computers. it is even possilbe
that some class of positions one engine can do great but not another.
On the other hand, human expert with deep knowledge, and have the access to the
engine can really make some good book.

2. it is hard to keep up with the latest development in the openning theory. For
my chinese chess program, more than half of the lost are directly from this. for
chess this is better because the new games are put online pretty early, but it
still a time comsuming and never end process, unless the book builder become
automated.

This is the weakest link in my chinese chess program. Because i am in US, away
from most chinese experts, and my auto-generated is far from satisfy.

Ren.



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.