Reply
 
LinkBack Thread Tools Display Modes
  #1   Report Post  
Old July 22nd 13, 09:59 PM posted to rec.games.chess.computer
external usenet poster
 
First recorded activity by ChessBanter: Jul 2013
Posts: 1
Default Engines for error categorization?

Hello, is anyone aware of any chess engines that attempt to categorize / explain errors?

I am hoping there is a chess engine that can be used as part of an automated or semi-automated coaching system. The classification of errors in a player's past games could be useful for this purpose, as discussed in several popular books and articles, some of which are mentioned he http://exeterchessclub.org.uk/content/assess-your-chess

A chess beginner would be forgiven for being overwhelmed by advice. There are so many different kinds of chess study one may engage in. How can you choose? Some say "just play more!" others say "just study tactics!" but really whose advice should carry the day?

Ideally there would be a system that would look at my last 20 or 200 games and give me some personalized advice. In my case it would say "your biggest problem is simply leaving pieces hanging. Fix that and you can expect to gain about 100 ELO points". And then game by game I can see a chart showing my progress in Hanging Piece Blunder rate, and once I've made enough improvement then the system tells me what I can/should work on next.

I don't expect such a system to be useful over some ELO level, perhaps not above 2000. But for terrible 1100 players like myself, it's a whole different ballgame.

What I'm asking for isn't simple, I know. I'm an experienced software engineer, and ages ago I wrote a simple alpha/beta pruning engine for a different game (Hnefaftl, http://en.wikipedia.org/wiki/Tafl_games).

But improving upon the current classification of errors seems doable -- since currently all one gets is the number of centipawns!

To get back to the point, are you aware of any engines that give some structure or color on the nature of errors?


Thanks in advance for any thoughts! --David
  #2   Report Post  
Old July 25th 13, 08:10 AM posted to rec.games.chess.computer
external usenet poster
 
First recorded activity by ChessBanter: Jul 2006
Posts: 1,015
Default Engines for error categorization?

On 22/07/2013 21:59, David Joerg wrote:

Hello, is anyone aware of any chess engines that attempt to categorize / explain errors?


It wouldn't be for the engine to do that so much as the GUI. Later
versions of Fritz have acoaching mode that might do what you want but
really there is no substitute for letting the computer annotate the game
(or blundercheck it if you don't like it's naff comments) and then
playing through looking carefully at any sudden score discontinuities.

I am hoping there is a chess engine that can be used as part of an automated or semi-automated coaching system. The classification of errors in a player's past games could be useful for this purpose, as discussed in several popular books and articles, some of which are mentioned he http://exeterchessclub.org.uk/content/assess-your-chess


Could be done but I don't know of anything that actually does it that way.

A chess beginner would be forgiven for being overwhelmed by advice. There are so many different kinds of chess study one may engage in. How can you choose? Some say "just play more!" others say "just study tactics!" but really whose advice should carry the day?


Initially learning to solve chess puzzles and working on the basics a la
TASC (which is very long in the tooth now but still very good) will get
you a long way. If you are down at the level where you often hang pieces
then training to look at all the possibilities and counter moves will
generate a huge increase in performance for minimal effort.

Ideally there would be a system that would look at my last 20 or 200 games and give me some personalized advice. In my case it would say "your biggest problem is simply leaving pieces hanging. Fix that and you can expect to gain about 100 ELO points". And then game by game I can see a chart showing my progress in Hanging Piece Blunder rate, and once I've made enough improvement then the system tells me what I can/should work on next.


TASC and at a more advanced level CT Art 3.0 does this by setting ever
more difficult puzzles and requiring you to solve them. It can deduce
from the mistakes you make what needs improving.


I don't expect such a system to be useful over some ELO level, perhaps not above 2000. But for terrible 1100 players like myself, it's a whole different ballgame.


CT Art is probably most useful for that 2000+ ELO range.
TASC for 1000-1600. I don't think there are many engines ideal for
training club level players optimally today as maximum strength sells
and they are relatively poor at simulating lower ELO play with human
like errors. They play to get the right win/lose ELO statistics but the
errors are often too crass and easy to spot in the range that I play.

What I'm asking for isn't simple, I know. I'm an experienced software engineer, and ages ago I wrote a simple alpha/beta pruning engine for a different game (Hnefaftl, http://en.wikipedia.org/wiki/Tafl_games).

But improving upon the current classification of errors seems doable -- since currently all one gets is the number of centipawns!


Well you get that and if you ask the engine to provide best line
annotation when there is a mistake you can get an idea. But quite
honestly if you want to improve there is no alternative but to play
through the game with the computer in "Infinite analysis" mode and see
how your (and its) moves rank against the machines idea of best play.

In modes played at fixed ELO the engine will also make its own mistakes
characteristic of the level of play.

You need to set the window for annotation to be appropriate to your
level of skill. Mine is set to 30cp today but from your description you
probably want to use a window of 150cp until you can mostly avoid
hanging pieces. I think the default was 50cp but that will produce too
much flack for a beginner - you need to see and work on the big errors
first and then do stepwise refinement.

To get back to the point, are you aware of any engines that give some structure or color on the nature of errors?


Not really. Any strong engine will do multiline analysis that will let
you see the ranking of the move you chose and the best play.

I think this would be a feature of the GUI rather than the engine.

You could probably use Chessbase to glean some statistics from your own
games but how effective it would be as a training aid I don't know.


--
Regards,
Martin Brown
Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Endgames & Engines Martin Brown rec.games.chess.computer (Computer Chess) 5 March 9th 07 01:52 PM
Fritz 10 FRC option is greedy. Won't allow other FRC engines. DorothyFan1 rec.games.chess.computer (Computer Chess) 0 December 1st 06 03:05 AM
The cheating IBM Wlodzimierz Holsztynski (Wlod) rec.games.chess.politics (Chess Politics) 148 July 22nd 06 02:52 PM
The cheating IBM Chess One rec.games.chess.misc (Chess General) 120 July 22nd 06 02:52 PM
The cheating IBM Chess One rec.games.chess.computer (Computer Chess) 100 July 22nd 06 02:52 PM


All times are GMT +1. The time now is 07:20 PM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
Copyright 2004-2019 ChessBanter.
The comments are property of their posters.
 

About Us

"It's about Chess"

 

Copyright © 2017