Winboard Forum

Hi,

is there a standard format for opening books? I would like to add a book to my engine, and I don't want to reinvent the wheel

Thanks,

Giuliano

No. Everybody makes their own.

Typically, you read a big pile of games between very strong opponents (I like to use SSDF top rated games, OTB 2700+ players and correspondence GMs) and then filter them in the way that you like.

The experts will use things like BCO, MCO, ECO, etc. to make really good move choices. I am too lazy for that. But clearly, you will get a better book if you do it.

I suggest using a database like FastDB if you are a database guy.
Most people resist that direction for some reason, although it seems very natural to me.

Dann Corbit wrote:No. Everybody makes their own.

It's a pity

Typically, you read a big pile of games between very strong opponents (I like to use SSDF top rated games, OTB 2700+ players and correspondence GMs) and then filter them in the way that you like.

The experts will use things like BCO, MCO, ECO, etc. to make really good move choices. I am too lazy for that. But clearly, you will get a better book if you do it.

Ok, thanks. Incidentally, I've heard about your game collection: is it still available?

I suggest using a database like FastDB if you are a database guy.
Most people resist that direction for some reason, although it seems very natural to me.

Thank you for this interesting suggestion, I think I'll go towards this direction

Giuliano

Hi Giuliano,

Dan Honeycutt (Bruja) kindly offered his opening book code
(with book) for any engine author who needs it.

http://wbforum.volker-pittlik.name/view ... ight=bruja

GES engine (David Weller) uses this.

regards,
Jim.

Bruja book's on Dann's site
(Hope you don't mind direct link Dann)

http://cap.connx.com/chess-engines/new- ... book11.zip

Jim.

Giuliano Ippoliti wrote:

Dann Corbit wrote:No. Everybody makes their own.

It's a pity

Typically, you read a big pile of games between very strong opponents (I like to use SSDF top rated games, OTB 2700+ players and correspondence GMs) and then filter them in the way that you like.

The experts will use things like BCO, MCO, ECO, etc. to make really good move choices. I am too lazy for that. But clearly, you will get a better book if you do it.

Ok, thanks. Incidentally, I've heard about your game collection: is it still available?

I suggest using a database like FastDB if you are a database guy.
Most people resist that direction for some reason, although it seems very natural to me.

Thank you for this interesting suggestion, I think I'll go towards this direction

Giuliano

My chess stuff is on http://cap.connx.com

This is a compressed PGN file with mostly good games in it:
http://cap.connx.com/chess-engines/new- ... ew.pgn.bz2

If you want a hand with the DB approach, let me know.

Jim Ablett wrote:Bruja book's on Dann's site
(Hope you don't mind direct link Dann)

http://cap.connx.com/chess-engines/new- ... book11.zip

Jim.

I do not mind. The share is there for everyone to do what they want to with this stuff.

Since my program is in Java would it be ok if I rewrote it to work in Java?

Hi Jonatan

Since my program is in Java would it be ok if I rewrote it to work in Java?

Maybe easier to check out some already written Java opening book code.
There are a few Winboard Java chess engines (with source) that use their
own opening books >

JChess - by Tomasz Michniewski
http://www.tytanchess.republika.pl/

Olithink 412j - by Dr. Oliver Brausch
http://home.arcor.de/dreamlike/

Olithink reads a pgn file as a book, so no
conversion needed.

I'm sure if you ask the authors nicely they'll let you use
the code in your program.

regards,

Jim.

Ah, I'll check those out. Thanks.

Giuliano Ippoliti wrote:Hi,

is there a standard format for opening books? I would like to add a book to my engine, and I don't want to reinvent the wheel

Thanks,

Giuliano

Hi Giuliano,

Wing's opening books are derived from 160,000+ top GM games, mainly from the period 1950-2004, stored in a 100 MB pgn file and it contains 7,457,216 moves in total.

My own created opening books are binary files that contain 4 sections:
1. a header, containing general information, a unique signature, version number, number of positions stored (nrpos), etc.
2. nrpos positions, sorted and stored as 64-bit Zobrist' keys.
3. nrpos+1 starting addresses, for all reply moves, these addresses refer to section 4 below:
4. all reply moves with frequency of occurance.

I don't know how others do it, this is probably not the most optimal way, but replies from the opening book are usually looked-up in about 30 milliseconds or so.

Again I think we all could benefit from an 'open standard'.

Stef