Is It possible to Design this kind of Software?

[swami]

If so, how much time of work one needs to invest on it? I'm no programmer but I'd like to know what you think about this:

* Get a Pgn, look for certain move in a game where the elo jump shows a factor of > +2.0 from move number x to move number x+1.

* Save that position to the database, as position number n along with the solution.

* Go to the next series of moves in the same game and look for the evaluation variance. Especially keep looking for moves where the eval jump from one move to another shows up +1.5 or +2.0 score and save them to the database as the next sequence of problems. (n2, n3....)

* For example, In a game, Computer playing black side showed the eval score of +0.23 in move number 15 and jumped to +2.24 on move number 16 after opponent made a mistake. This one gets saved into database. And in the same game, the computer shows +2.34 on move 21 and in the next move its eval lept to 4.14, this position gets saved to the database as position number 2 along with the solution.

* After searching through the game 1 from the pgn and saving certain positions to the database, the "program" or a tool, loads up game 2 and does the same thing all over again until the end of the last game on the pgn.

We get series of tactical positions from a single pgn. This could make an useful tool. There are so many engine games, this tool could search through all the games and filter the ones that are tactical in nature in terms of evalution rise. This could prove really helpful to human trying to learn tactics. We could gets millions of tactical problems in one simple search via this application.


Or is there already such tool available elsewhere? I know you'd say its a "blundercheck" option in chessbase or Arena that does this. But it doesn't exactly do what I'm suggesting here.

[Dann Corbit]

If so, how much time of work one needs to invest on it? I'm no programmer but I'd like to know what you think about this:

* Get a Pgn, look for certain move in a game where the elo jump shows a factor of > +2.0 from move number x to move number x+1.

* Save that position to the database, as position number n along with the solution.

* Go to the next series of moves in the same game and look for the evaluation variance. Especially keep looking for moves where the eval jump from one move to another shows up +1.5 or +2.0 score and save them to the database as the next sequence of problems. (n2, n3....)

* For example, In a game, Computer playing black side showed the eval score of +0.23 in move number 15 and jumped to +2.24 after opponent made a mistake. This one gets saved into database. And in the same game, the computer shows +2.34 on move 21 and in the next move its eval lept to 4.14, this position gets saved to the database as position number 2 along with the solution.

* After searching through the game 1 from the pgn and saving certain positions to the database, the "program" or a tool, loads up game 2 and does the same thing all over again until the end of the last game on the pgn.

We get series of tactical positions from a single pgn. This could make an useful tool. There are so many engine games, this tool could search through all the games and filter the ones that are tactical in nature in terms of evalution rise. This could prove really helpful to human trying to learn tactics. We could gets millions of tactical problems in one simple search via this application.


Or is there already such tool available elsewhere? I know you'd say its a "blundercheck" option in chessbase or Arena that does this. But it doesn't exactly do what I'm suggesting here.

There is this tool:
http://home.pacific.net.au/~tommyinoz/gameanalyser.html

But it only works on Winboard debug logs (as far as I know). It would be nice to have the same thing for Arena output and other tagged PGN output.

[swami]

Chess Hero comes close to doing this, but it does it in different way though.

I should talk to the author of Chess Hero and share some interesting ideas, to see if he can implement them in future releases.

[Dann Corbit]

There are tons of annotated files (e.g. 'games with comments' from here):
http://www.computerchess.org.uk/ccrl/4040/games.html

that could be fed into a "surprise discovery" engine to produce the positions you are looking for. I think it would be very useful also.

[Edmund]

There is this tool:
http://home.pacific.net.au/~tommyinoz/gameanalyser.html

But it only works on Winboard debug logs (as far as I know). It would be nice to have the same thing for Arena output and other tagged PGN output.

The latest Version (1.2) can also read Arena-Comments

[Tommy]

Hi Swaminathan,

Others have mentioned my tool already (Game Analyser), but one thing I haven't advertised on my web page is that the tool can export positions that it finds to an EPD file. This maybe a more convenient format for you to work with and maybe to convert to a database format of your choice.

Cheers,
Tom.

[Matthias Gemuh]

Have you seen GM Moves of ChessGUI ?
Collects stuff only for on-going game, though.

Matthias.