Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A new chess program on the horizon?

Author: Uri Blass

Date: 14:01:48 11/15/01

Go up one level in this thread


On November 15, 2001 at 12:20:37, Christophe Theron wrote:

>On November 15, 2001 at 12:06:24, Slater Wold wrote:
>
>>I heard it through the grapevine that Microsoft is working on a chess program.
>>
>>Nothing like Deep Blue, but something to rival Chessmaster, and I am guessing,
>>Chessbase.
>>
>>With the money, resources, and brains at Microsoft, I think they have the
>>ability to make one of the best chess programs on the market.  They could afford
>>to get the best GM's and use some of the best programmers in the world.
>>
>>I was curious on what the thoughts of this forum is, on such an idea.  I know
>>most people here don't like Microsoft, but please keep in mind I am asking for
>>your input on a Microsoft chess program, not their business ethics.
>
>
>
>I guess that if Microsoft wanted to publish a chess program they would not
>gather together GMs and programmers.
>
>They would just buy an existing chess program.
>
>Time has shown that it takes a lot of time (years) to build a strong chess
>program, and that you do not need to involve grandmasters in the process.
>
>What you need the most is TIME.

I agree that you need time but I believe that you need clearly less time if a
lot of people are involved.

I agree that you do not need grandmasters but I believe that it is possible to
do something clearly better than the existing programs by defining selective
search rules.

You need first to define rules to decide if a move is illogical

Example for the rules that I have in mind is something like Bg2-h1 is not
logical if the move is not a capture and there is no bishop at h3 and maybe
another condition that I did not think about.

You need to analyze millions of games in order to discover rules and to check
that the rules  are always correct or at least correct in 99.99% of the cases
and for this task you need a lot of workers.

You need also a lot of programmers to develop the source code for thousands of
cases with no bugs.

You also need later to test how to use the selevctive search rules.
You can use selective search rule that prune all the moves that are considered
as illogical but it seems to risky and it is more logical to allow moves that
are considered as illogical moves if you do not have too many illogical moves in
the same line.

You also may need a lot of computers to test possible changes in the evaluation
function.

I believe that all this task can be done in one year if you have 1,000,000,000$
to waste about it but I also believe that microsoft is not going to do it.

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.