Benoit St-Pierre wrote:
Hi!
> 1. Can I use Scid to build opening books that can use chess engines ?
>
>>From what I gather, there is no standard way to represent opening
> books.
Right. Allmost every app uses its own format.
> I know that GNUchess uses a simple .pgn file.
This is inefficient for large books. Its inefficient to
search and wastes a lot of disk space. Just compare what you
gain if you transfer from pgn to scid format.
> That seems to imply building an opening database, i.e. a
> collection of game fragments representing a more or less
> complete repertoire.
Yepp.
> Does anyone tried to build an opening database like that ?
AFAIK Arena's Mainbook is built that way. And you can do
something like that by book tuning window.
> Building a .pgn book with Scid is a bit unintuitive, since PGN
> databases are read-only.
Eh?? Of course you'd just set up a usual scid database and
just export it to PGN for book building. PGN would just be
an intermediate format. You've to resort to PGN only cause
most apps that build binary books only know how to read PGN.
The nice thing of PGN is, that it is a standard. There is no
real necessity to use it as input for book building.
> But that's the most expedient way, to my knowledge. Is
> there a speedier way ?
AFAIK not today. This would require Scid itself to write the
book. In principle, however, it should be possible to cut
the PGN-Export - polglot PGN Import way short and give the
data directly to polyglots internal routines.
> 2. How to use .bin books, and how to make one using Polyglot ?
>
> This possibility may be a sad necessity, sad because Scid
> could very well be used to build opening books. Most of
> OTB players use databases for that : why not use these
> efforts to build up books ?
Hardly anybody will give you a hand weeded opening database.
;)
> It's sad, but not so bad : Polyglot does not seem that
> tough to master for someone who prefers to build books
> than buy some.
Its actually a pretty simple commandline call of polyglot
against a PGN file. (Again, PGN only as polyglot only knows
how to read this format. In principle, it can be any other
format.)
> Anyway, does anybody know how to use Polyglot ?
RTFM ;) That is, it's written in the docs of Polyglot of
course. See it's accompaning readme.txt. Basically:
polyglot make-book -pgn input.png -bin output.pbk
(Note: polyglot books do not have to have this absolutely
silly .bin extension, and I'd strongly vote for changing it
in Scid as well...)
Additional parameters allow you to define the maximum number
of plys for the book, the minimum number of games a line has
to appear, to build a book for white or black only, or to
set the probabilities for all moves from a given position to
the same value.
Drawback of the above: for a simple minded approach one
would just use a database of grandmaster games. Say, your
referenc base above 2400 ELO or whatever. If this gives a
large PGN you really need a HUGE ammount of memory to build
a book. I did this once for, if I remember correctly, all
games above 2500, and it was only possible on a box that had
1.5G of _FREE_ RAM. (I admit that I missused one of our
compute servers on idle time. ;)
> it seems that a .bin book is simply a .pgn file compiled by Polyglot.
Sort of. It contains statistics about the moves and those
are also adopted based on engine experience.
> Is this source good enough ? Can some experimented user explain the
> process of building a .bin book in a simple manner ?
I fear, as for me a command line is a native interface I'm
the wrong to do this ;)
BUT, given a polyglot executable at hand it would be very
easy to write a simple GUI for it. However, I think doing
it directly from the DB without dumping a PGN first would be
preferable.
> 3. How one can build a good opening book by looking at a
> well-chosen database ?
See above. Just dump it to PGN and run polyglot against it.
> But what would be a well-chosen database for you ? I plan
> to study chess more thoroughly, but I can't decide how to
> organize my chess databases. Can you provide me with some
> user cases ?
I think, to build a complete book is quite a work load. But
you can of course add the lines you studied to a base, then
set -min-game to 1 for the building process and it would be
sufficient to add all lines as individual PGN files in your
base. That is, it's format would be that of Pascals initial
opening trainer input base. You'll find a hopefully
enlightening commentary on it's format in the F1-help.
> (What means locking a database and selecting a fixed font ?)
? What are you refering to?
> 4. Why not use the repertoire editor ?
> It seems to me that this window was meant for this.
Yepp.
> At least, when it comes to search incoming game files,
> e.g. TWIC. But it also seems very underused.
Yepp.
> Does anyone use this editor ? If yes, how, and if not,
> why ?
I did heavily while back, it is still one of the best ways
to store a larger number of predefined queries. However, the
mask for the tree window are much more powerful when it
comes to handle transpositions. Simply, cause the masks are
based upon a position and not about a line of moves. Once
you added your repertoir to a mask and you go to a specific
position you've all games in the games list and the best
games sorte out in the best games window. Here it is much
more flexible. The only thing the Masks IMHO currently lack
is a quick way to jump to certain positions, ie. to really
use the markers beyond just notification. E.g. ther're
markers for "Bookmark" or "To Study" which are very nice,
but equally well hard to find again as you can not just get
a list of all your bookmarks or postions to study and just
jump to them by a click with the mouse. If I got Pascal
correcly, he wanted to add this later. I'd also find it
useful, if one could also get a "line representation" of a
mask, that is make a classical opening base out of the masks
contents. Pascal told us that there'd be some ambiguities
here and if I got him correctly he has it on the todo
somewhere in the "non pressing" area. This thing however
would be most usefull for your book building idea.
> Thank you very much in advance for your help,
Hope that helps a bit.
--
Kind regards, / War is Peace.
| Freedom is Slavery.
Alexander Wagner | Ignorance is Strength.
|
| Theory : G. Orwell, "1984"
/ In practice: USA, since 2001
------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword
_______________________________________________
Scid-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scid-users
