Author: Christophe Theron
Date: 12:06:33 06/22/00
Go up one level in this thread
On June 22, 2000 at 13:22:20, Fernando Villegas wrote:
>Hi:
>Maybe the issue could be approached not just adding new code lines to the same
>cluster of code lines -and so paying a price, as you say, in other aspects of
>the gaME- but making use -how many times I have adviced this thing and how
>increasingly sure I am that i am in the right track!- of an specific module to
>tell the program what to do in such a case, with its own cluster of code lines
>for that situation. I have said it once and again: a strong human player never
>makes calculations before trying to grasp what kind of calculations he must go
>for. This thing of adding or substracting pieces of code of just one, unique
>code, seems to me somewhat primitive. Every time you put or you substract
>something in order to face a problem, the entire code is affected. In my view
>-sorry if I talk of something that I do not practice- the program should be an
>elastic things, something that arrange himself according the situation, that is
>to say, that puts or live outside some piece of code that are useful or unuseful
>for every ocasion. In the case of blocked pieces, clearly some special cluster
>of the entire code should be activated and the rest desactivated because
>different rules rule here. . Yeah, yeah, how to develop the super module to say
>what must be done? Years of work I suppose, but that is the track, I bet.
>Sorry for your time
>Fernando
Actually what you are referring to is called "preprocessing". The program,
before it starts enumerating all the possible combinations up to a given depth,
takes a good look at the position and tries to extract some key features of the
position and "configure" itself so it is going to use during its search the
appropriate bits of knowledge.
For example, if you see that you are in an endgame position, it is obvious that
no opening knowledge is needed. But it does not apply to everything in the game.
I doubt it is possible to apply this to blocked pieces. Actually I KNOW it is
not possible because I have already tried in the past. It is not possible to
reliably identify blocked of future blocked pieces by just looking at the "root"
position. A blocked piece can be freed by a combination of moves, while a
blocked piece can appear at a strange place deeper in the tree.
This is exactly, and unfortunately, the kind of knowledge that must be applied
to all the positions that are seen by the program. And it is unfortunately a
very expensive term to evaluate.
Christophe
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.