Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The Book Problem, Open Book Format proposal, Request for comments.

Author: Dann Corbit

Date: 15:21:56 03/03/06

Go up one level in this thread


On March 03, 2006 at 17:14:26, Odd Gunnar Malin wrote:

>On March 03, 2006 at 16:46:13, Dann Corbit wrote:
>
>>On March 03, 2006 at 16:24:49, Odd Gunnar Malin wrote:
>>
>>>On March 03, 2006 at 13:59:27, Jay Urbanski wrote:
>>>
>>>>I'd like to get some comments on an idea that's been growing in the back of my
>>>>head for some time now.  The problem is that of opening books.  We now have two
>>>>open protocols for communication between engine and GUI with Winboard and UCI -
>>>>but every GUI has chosen to go the route of a proprietary book format -
>>>>presumably to lock users into that GUI.  (Even Arena does this, something that
>>>>puzzles me since it is free)  This may be good for the vendors business models,
>>>>but it bad for users since it limits choice.
>>>>
>>>>So what is the engine author to do?  They can:
>>>>
>>>>A.  Pick a GUI to support (and alienate users of other GUIs)
>>>>B.  Make an engine book (with almost certainly less features than the GUI books)
>>>>C.  Support all the GUIs (way too much effort, and almost certainly inconsistent
>>>>quality of books)
>>>>
>>>>All of these have their shortcomings, so my proposal is the establishment of an
>>>>open book format.   The idea is that with one format option "B" could be a lot
>>>>more attractive, especially if enough features were put into the book
>>>>specification.
>>>>
>>>>A common objection to this idea is "It won't fly without support from the GUI
>>>>vendors, and none of them will support it".  I agree none of the GUI vendors are
>>>>likely to support it at first (except maybe Arena?) but I contend that if enough
>>>>engine authors support it, *and* we could get a decent book editor/viewer
>>>>written so that users could view the book, tweak it, etc - it wouldn't matter
>>>>whether the GUI authors support it.   And if it became popular enough, the GUI
>>>>vendors might add it as a feature due to customer demand.
>>>>
>>>>So I'd like comments/feedback/criticism from everyone.  Book authors especially,
>>>>what features would need to be present to make it a good format?  As a user I
>>>>would defnintely want book learning and an intuitive but flexible syntax to
>>>>weight moves.
>>>>
>>>>Would an existing format like Polyglot be a good starting point?  Would the
>>>>Arena team consider donating the .ABK format as the basis?  Would anyone be
>>>>willing to donate programming skill & time toward a decent book editor/viewer?
>>>>
>>>>Am I the only one who cares about this?  :)
>>>
>>>Isn't it easier to agree on some import rules for importing pgn-files, eg. !, ?!
>>>etc. after the move are interpretet as the same in all gui when importing.
>>>ChessPartner have another system that also could be considered, it allow you to
>>>specify a score for each line in the pgn file.
>>>
>>>The problem I have is that I create my books as games with variations in a DB
>>>program (CA for me). Then I have to make all sorts of converting to be able to
>>>import it into the GUI I want to play with, ChessMaster, Shredder, ChessPartner
>>>or even Fritz. I'm maybe excentric but I usually use the GUI that was attended
>>>to the engine when I bought it.
>>>
>>>Maybe there is some shortcomming for the few who run engine-engine matches, but
>>>for the big bunch that use the programs main feature it is realy enough with the
>>>few scoring you get from a pgn-file (??,?,?!,<none>,!,!!). If the GUI in
>>>addition have the ability to import Black and White into a repertoire for the
>>>engineplayer so you can use a single book for both color when training, I think
>>>all may needs should be taken care of.
>>>
>>>Odd Gunnar
>>
>>Suppose that a move is marked !! but your engine loses when choosing that move
>>ten times in a row (a very real possibility if it is a good move but your engine
>>does not know the plan).  Wouldn't it be nice to have a book format that
>>understands statistical inferences and knows it ought to switch to something
>>else?
>
>Ten game in same line? wouldn't that take a year to come tru if the book have a
>few lines. If the !! are missplaced you can easely correct it and import it
>again to each gui.
>
>If you realy want to practice against the line the engine lose, why want you the
>gui to avoid this? Neither the gui or the engine have any soul so I don't think
>they care.
>
>Else I'm a fan of a good position learning, designed to give variation to the
>player even if he start from a middle- or endgame position.
>
>I undestand from the other post that the main focus is on engine-engine games,
>but for me an engine is only a tool for the human-player to train or get some
>relaxed games from.
>
>Some scenary:
>You want to create a thematic tournament and need some practice, then you can
>create a training-book without thinking what gui the players prefere to use.
>
>Your team will play against the next opponent in the league next week, then as a
>captain, or trainer, you want to prepare the players for this match.
>
>You have study an opening and want some practice on this.
>
>You will enter a tournament and create a trainingbook for this based on the
>repertoire of your expected opponents.
>
>The !, ?? etc. is to have some control of how often each line are selected,
>maybe one opponent play 1.f4 and 5 usually start with 1.e4, of course you want
>the engine to play the 1.e4 line mostly because there would be more lines in
>there.

What you suggest is also fully supported by the model that I proposed, which has
provision for NAGs.



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.