Author: Graham Laight
Date: 06:57:32 07/04/00
Go up one level in this thread
On July 03, 2000 at 11:44:17, Mike Saavedra wrote:
>There are various reasons why people buy chess software, but for me, it was to
>improve my game, more specifically, to annotate my games. I own Fritz 5.32 (love
>the program) but was kinda let down by the annotations feature. I am a Club
>level player and would like to have a program to have the knowledge, and the
>ability to suggest moves in quiet positions, AND give me an explanation WHY it
>is better using established chess principles. Fritz is great for telling me
>tactical errors, but what of the rest of the time? ("Explain all moves" feature
>on Fritz is too simplistic). Chess Mentor is a program that approximates what I
>am talking about, yet the content is preprepared, and will not show me the
>"error of my ways" in context of my own games. Maybe look at ALL my games and
>notice tendencies, and chronic problems I may have overlooked in my own games.
>
>Is it possible? Teach a program the Opening, Middlegame and Ending principles of
>some classic books, and applying it to teach me how to improve my game using my
>own games? It does not have to be GM strength! A GOOD Master level annotator
>will do nicely for a patzer like me! :)
>
>I am thinking of purchasing Junior, hopefully it will give me better positional
>alternatives to the moves I have made than Fritz. If you think there is another
>program that annotates better, please let me know. Thanks
>
>Mike
Unfortunately, the prevailing doctrine in computer chess is to write programs
that evaluate huge numbers of nodes in the game tree quickly, and then select
the move which, on the basis of these node evaluations, leads to the least
possibility of ending up in a bad position.
In effect, the computers have no idea why they pick a move - and it takes a lot
for their programmers to understand what's going on as well (how long would it
take for a programmer to comprehensively analyse even 1 second's worth of work
on a 1 GHz computer?).
To be able to analyse a static position well, a methodical system of organising
knowledge would be required (most programmers, as far as I can tell, do not
organise knowledge methodically). Then, the pieces of knowledge which are having
the greatest impact on the evaluation could be picked out.
But in any case, what is an annotation? It's the piece of knowledge that the
expert feels is relevant (or interesting) in the current situation. But who's to
say that they're right? Or that there aren't other pieces of knowledge that
aren't also highly important in the current position?
What's annotation all about, really? If we're honest, isn't it really just
something to give a feeling of comfort to the reader?
This problem doesn't just apply to chess annotation, it applies to sports
commentary, financial advice, political comment - lots of things really. It's
just that when you want to talk about generating this stuff automatically,
you're forced to think about these things!
-g
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.