Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Algorithms vs. knowledge - What to do next?

Author: Marc van Hal

Date: 02:51:44 06/04/02

Go up one level in this thread


On June 04, 2002 at 03:34:19, Russell Reagan wrote:

>Lucky for me I already own all of those books you mentioned, as well as about 70
>others ;) I have tons of ideas for evaluation already. That's the stuff I've
>been dreaming up for years. I think my only problem with my evaluation function
>is going to be getting it to evaluate the way I want efficiently.
>
>Thanks for all of that info. It'll definetly keep me busy for quite a while :)
>
>Russell

But first look at this you only can solve the problem if you know what the
problem is!
Ok I know the that even if you know the problems it can not be answered by
A simple answer because mostlikely it wil be a mix of things.
I think it would be great to look at position where all programs do have
difeculty's with.
Strangely enough one of them is the fight for the centre.
And the fight for time.
An other problem is that most problems are tweaked in such away that they
actualy prefer to play Clasical openings which does not match with the previous
statement.
With the exeption of Junior
Even here you can get some good tweak points just for starters the queens
gambit.
Fritz7 still wants to play for the keeping of the c4 pawn.
It can be intresting to see what you have to change in your settings
that it will see that this is not a good idea.
In other positions a move order like g6.Bg7 f6 can be good to get an attack on
centre pawn on e5 programs of today will not likely play this because of
kingsafety settings.
(When I started with analysing with programs g6 alone was already not an idea a
program would think of because of the same reason most of the time if not in the
openingsbook all pawns before the king stayed if posible on it's place.)
Just look at the showdown of Gulko GambitTiger

This kind of coments you should not see as an insult but rather as an option to
improve the programs
It is also so if you have solved all this you also have solved the
the way to play against the Stonewall.
But talking about Junior
It's learning actualy only works fully aslong as you don't turn of the program.
Is there not a better way to store this learning in a file?
I even could give a simple answer on how to improve Rebel personelety's
But I can't give an answer on how to make it play perfectly.
Though it can come prety close to it.
I admit that even my chess knowledge is to litle to answer that question.
Then again where are we talking about then the 80'th release of progam which has
solved chess?
Just to show that good things sometimes can have a bad side affects.
Or the Dream from now can end up in a nightmare of the future.

       (C)06-04-2002 Marc van Hal

Hehe



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.