Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: =about Tempi, pawn moves and eval function -(i dis agree with M>)

Author: Ingo Lindam

Date: 05:09:17 07/06/03

Go up one level in this thread


On July 06, 2003 at 05:00:16, Andrei Fortuna wrote:

>That's mobility. To quote Mr. Hyatt's opinion on this subject "is mobility the
>'cause' of a good position or is it simply the 'effect' of a good position? I
>believe the latter is closer to the truth. Otherwise, moves like a4 would
>be _good_ moves because they instantly improve both the real and potential
>mobility of the a1 rook."

Well, I guess a4 is a good move in a huge amount of positions, in particular
when there is a rook on a1. In some of them there might be a even better move.
Ofcourse whether a4 is a good move depends on a lot of features. I wonder if the
above question is of great interest in computerchess. The tasks of making good
moves (and thereby changing the features of the position) to get into a good
position and that of recognizing good positions by a number of suitable features
seems to be the same.

>Easy to say. But what games to include to make this database ? Even GMs make
>some blunders in their games and opening theory changes (a bit) in time.

If you have a lot of games played on a high level there should be not to much
examples of the same or similar (positional blunders) leading to more wins than
losses.

>Even if you make this good squares table based on openings : you are still
>saying that in some openings some squares are just very good for some pieces,
>and I disagree with this because it doesn't account for the placement of other
>pieces.

I think, too, that it might be dangerous to learn features like where to place
best a single piece just depending on the opening. But I am sure it would be
worth to know what pattern of placements of (some of) the pieces is successful
against the pattern of placements of (some of) the opponents pieces. And it
might to to often depend to much on the (exact) placement of the opponents
pieces, in particular in closed positions, where sometime knowing an adequate
plan and where to place the pieces is much more worth than to calculate some
hours in the fog of 20 or more plies. (Although myself would tend to do so and
loose on time)

>For example what in an opening 'X' it is found that knight's position on f3 is
>good, but this because in the analysed games the knight protects the pawns on d4 and e5, and in your game you place the knight on f3 but there are no pawns on d4 or e5 so it doesn't have the same role as in that opening. Maybe this is not a very good example but it does show what I mean above.

You are very right at this point in my eyes. Whether there are pawns on d4 and
e5 is obviously a very important feature in a position not only in question of
where to place the knight. The evaluation of good squares, good moves and good
plans by using the knowledge or experience of a lot of high level games should
not lead to judge the place of the knight on basis of the very first moves on
the games, but on the basis of the featers of the current position. In a
position where pwans are places on d4 and e5 the machine should evaluate the
position of the knight (and some other pieces) much more on the basis of
positions and games with pawns on d4 and e5 than on positions occured after the
same opening.

Internette Gruesse,
Ingo



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.