Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: AI for human-like chess analysis and annotation

Author: KIRK SEARS

Date: 05:34:44 03/22/03

Go up one level in this thread


VERY INTERESTING AND ILLUMINATING COMMENTS. THANK YOU.  I AM GLAD OTHERS ARE
CHALLENGED BY THIS AS WELL!


On March 21, 2003 at 15:15:35, Matthew White wrote:

>On March 21, 2003 at 12:40:45, Cliff Sears wrote:
>
>>Chessbase and Chessmaster it appears to me have to closet thing to having a
>>Grandmaster analyze your games.
>>
>>Anyone attempting to create a game analyzer that would annoatate your games like
>>a Purdy, Euwe, Nunn or a Chernov? :)
>>
>>That would be a very interesting AI programming task to be sure. How would this
>>even be attempted? What would be the algorithm.
>>
>>Everyone is trying to make a better chess playing engine but not a better chess
>>annotator
>>
>>Just throwing this out there :-s
>Part of the problem that would need to be solved is that chess engines don't
>really "understand" positions. That would need to be solved for an effective
>annotator. However, solving that would create a better chess playing engine
>(IMO, of course), so it becomes a chicken-and-egg problem. I have sought the
>same thing. It would really be nice if an engine annotator would award double
>exclams (!!) for a simple pawn move, as human annotators are capable of.
>However, I have only seen double exclams in machine-annotated games for queen
>sacs that lead to mate.
>
>Fritz hazards a guess at what the player was intending by the move (i.e.
>Demolishes the pawn structure!), though this is not always the POINT of the
>move. I played a game the other day in which I made a tactical shot that won my
>opponent's queen for a knight. However, when I analyzed the game with Fritz, I
>was disappointed to see that familiar "Demolishes the pawn structure" note,
>since, though I did take a pawn with my knight, the real point of the move was a
>discovered attack, not a destructive sacrifice.
>
>In any event, if you ever find a better analysis program, be sure to let us know
>;).
>
>Regards,
>Matt



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.