Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Na3 wins again - Deep Shredder played an bad move!

Author: Uri Blass

Date: 09:33:04 06/28/01

Go up one level in this thread


On June 28, 2001 at 10:55:21, Robert Hyatt wrote:

>On June 28, 2001 at 09:16:47, Uri Blass wrote:
>
>>
>>The book does not include every legal move so practically you need clearly less
>>than 197000 moves to cover 4 plies in your book.
>>
>>Another point is that some of the positions are the same with different order of
>>moves and you only need to rememeber the positions
>>
>>It is possible to cover 6 plies and it is practically enough to avoid repeating
>>the same line gain  and again.
>
>This is one of those cases where it "looks easy" but is not.  Try to sit down
>and write such a book and see how hard it is to type a few hundred thousand
>moves and make sure you don't include moves that lose quickly.

The computer can do it automatically and find some moves that do not lose
immidiately.

If it needs to analyze million positions for that purpose then it is possible to
do it by analyzing every position for a minute.

It is possible to do it in a few monthes by using some computers for this
purpose.
>
>Then the book learning is going to make you try each choice for the computer
>and then nix it when it loses.  Meanwhile, the same theme arises over and
>over with the computer out of book.  If it has an opening weakness, it will
>be obvious...
>
>
>
>>
>>The computer can avoid repetition of the same 3 moves in hundreds of games
>>when it has 5-10 responses in almost every position.
>>Even if you use only 2 plies you can get a lot of different games.
>>
>>first game may be 1.e4 c5 2.Na3 d6
>>second game 1.e4 c5 2.Na3 e6
>>third game 1.e4 d6 2.d4 Nf6
>>4th game 1.e4 c5 2.Na3 Nc6
>>5th game 1.e4 e6 2.Qe2 e5...
>>
>>You are going to  need a lot of games to get one repetition of a line that was
>>played in the past and humans have better things to do then to learn hundreds of
>>lines against the machine and they need not only to learn hundreds of lines but
>>also to find hundreds of lines to beat the machine.
>>
>>Uri
>
>
>You don't necessarily have to repeat the same "game" to win over and over.
>The computer is out of book so quickly the same weakness will be seen repeatedly
>since the book is not really helping it when it is that shallow.

The point is that the opponent will have to think in the game and simply
repeating is not enough so it is going to make it impossible for 2100 players to
win again and again or to win by repeating the same opening every game with the
idea of trying to correct the error in the next game.

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.