Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Real Text Chess game analysis

Author: Michael Henderson

Date: 11:15:32 01/31/05

Go up one level in this thread


On January 31, 2005 at 08:38:43, Philipp Bouillon wrote:

>Hi,
>
>I am currently developing a "Chess tutoring" software which is using the Crafty
>Engine as chess player. Now, I'd like to analyse any position with the engine,
>but in addition to the evaluation score of the position, I would like the
>program to automatically print out texts like: "Outch! This move was horrible as
>it pins the own Knight to the King. This leads to e7 which attacks the knight
>and wins it in the next move...".
>
>The problem with this is that I'd have to find out the "delta" between two
>subsequent positions (so: Is there a huge drop/increase in the position score)
>and then find out _why_ the score dropped.
>My plan is to look deeeeep into the evaluation function of the Crafty engine and
>try to map the scoring to pre-defined sentences. In my imagination (I confess: I
>have not yet looked at the sources of Crafty), I hope to be able to find things
>in the evaluation function like "at this point, the king safety is measured"
>and I hope to be able to "hook" my idea to this evaluation function.
>
>The text generation itself should be fairly easy (and customizable for different
>languages as well), the only problem I see is to link the sentences to the
>evaluation... So, before I delve deep into the sources: What do you think? Has
>this already been tried? Do you see any (more) problems?
>
>Thanks for your feedback!
>Philipp

How advanced and how good do you want the text output to be?  I'm sure you can
get it to print out the obvious stuff, but you cannot go any farther unless you
are a computer chess programmer.  You need to have complete control over the
search function in order to have complete control over the text analysis.  You
do not control the search, so you are limited by crafty.  Even if you could do
it, there are several more implementation problems.  In fact, you should ask
Hyatt about this.  If you could do it and it was beneficial, Hyatt certainly
would have made it already.

I recommend making _your_own_ program to do this.  There are no limitations with
that method.  Nobody can prove what your program will or will not accomplish.
It's all up to you.

Michael



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.