Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: why write a fast chess program ?

Author: David Dory

Date: 13:32:54 08/17/02

Go up one level in this thread


On August 17, 2002 at 15:17:32, Engin Üstün wrote:

>On August 17, 2002 at 14:14:08, Frank Schneider wrote:
>
>>On August 17, 2002 at 13:38:46, Engin Üstün wrote:
>>
>>>i want not to discuss about fast bitboards or bit operations!
>>>
>>>my title is why write a fast chess program,
>>>and not a selective search program like a human chess player.
>>
>>Even a selective program will be stronger if it's fast.
>>
>>>
>>>i positions if alpha > beta cuts the program, but if not is searching all moves.
>>>
>>>i am thinking about the program picked some 1-5 possible canditate moves in a
>>>position and search only them.
>>
>>This is a very old idea - in fact when computers were slow
>>everybody tried it (Kaissa, Chess 3.0, Pioneer).
>>
>>
>>Todays best programs are quite selective, they have a branching factor
>>of about 2.x. It's not easy to improve that, but of course everybody
>>tries.
>>
>>Frank
>
>i mean not null move , extensions or pruning of moves.
>
>i mean if the program try only 2 moves in a position that can search very deep
>and beats every human players.
>
>2^18 = 262144 nodes is enough :-)
>
>not thausends or millions of nonsense positions.

Believe me, Engin - they certainly tried. Chess programs were SUPPOSED to be
this intelligent AI program that would play chess very much like Grandmasters
do.

Unfortunately, after years of trying - the "intelligent" chess programs STUNK!

At the heart of the problem is we don't know exactly HOW a GM goes about
selecting the best moves to study. DeGroot and others have studied this
extensively, but it seems to be a skill that is almost rooted in the
subconscious - even the GM's can't tell you WHY they favor making this move
instead of that move, many times.

So a good working (abstract) model of how a human GM plays chess is still
unavailable. Lacking that knowledge, most AI methods for chess programs proved
ineffective. (Book learning worked out, though.)

The new "brute force" method has now become somewhat "smarter" (more selective),
and appears to be THE way for chess programs to beat the human GMs.

A simple computer can't solve a complex problem like chess the same way as our
complex brains, but it can solve the problem it's own way. :)

David



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.