Polyglot book : -uniform option

Discussions about Winboard/Xboard. News about engines or programs to use with these GUIs (e.g. tournament managers or adapters) belong in this sub forum.

Moderator: Andres Valverde

Polyglot book : -uniform option

Postby ChessKnight » 15 Jan 2009, 07:26

Can anyone describe the technical difference in using polyglot make-book with -uniform option and without using it?
ChessKnight
 
Posts: 48
Joined: 26 Dec 2008, 06:37

Re: Polyglot book : -uniform option

Postby Marc Lacrosse » 15 Jan 2009, 09:35

From a given position all candidate moves will be played with an equal frequency in case the book has been made with the "-uniform" parameter.

Marc
Marc Lacrosse
 
Posts: 116
Joined: 29 Jan 2005, 09:04
Location: Belgium

Re: Polyglot book : -uniform option

Postby ChessKnight » 15 Jan 2009, 10:15

Thanks Marc.
ChessKnight
 
Posts: 48
Joined: 26 Dec 2008, 06:37

Re: Polyglot book : -uniform option

Postby H.G.Muller » 15 Jan 2009, 12:38

Strange that you have to specify this duringthe making of the book, as it throws away information (on the move frequencies), and does not substitute anything in its place. This option would be much more logical on using the book. Then you can keep the information on the move weights in the book, but simply ignore it when picking a move.
User avatar
H.G.Muller
 
Posts: 3453
Joined: 16 Nov 2005, 12:02
Location: Diemen, NL

Re: Polyglot book : -uniform option

Postby ChessKnight » 15 Jan 2009, 13:07

H.G.Muller

Thanks for reply.

Can you please tell (explain) us the technical difference in performance between polyglot book made with -uniform option and one made without -uniform option?

Which will perform better (stronger)?

reg
ChessKnight
 
Posts: 48
Joined: 26 Dec 2008, 06:37

Re: Polyglot book : -uniform option

Postby ulver » 15 Jan 2009, 13:29

It is a matter of how your book will play an openings. uniform means that if you have 2 possible moves in book, i.e 1.e4 c5 or 1.e4 e5 (or 1.e4, or 1.d4, whatever) , both will have probability of 1, so if you created a book from selected 1.e4 games (only e4c5 and only e4e5) uniform option giving equal priority for both candidate moves, so your book playing e4c5 and e4e5. If your pgn has wide variety of openings resulting book may play e4e6, e4g6, e4Nf6, e4c6 and so on. Once again, it is a matter of what you want, wide book or only selected openings to be played.
ulver
 
Posts: 18
Joined: 13 Nov 2008, 09:52

Re: Polyglot book : -uniform option

Postby Marc Lacrosse » 15 Jan 2009, 16:48

H.G.Muller wrote:Strange that you have to specify this duringthe making of the book, as it throws away information (on the move frequencies), and does not substitute anything in its place.


It does substitute something : an equal frequency.
You do not take into account the fact that the most advanced way of making PG books is to build several books that you merge in a specific order.
For example if I just wish my engine to play either 1.d4 or 1.c4 as white and nothing else, I can make a one-ply long book with just two game fragments (first game is 1.e4 and second one 1.d4) with options "uniform" and "white-only". Then i use the book-merge function to combine this one with a second more usual book.

How could you replicate the way my engine will play with this combined book through an option that would be activated only during play?

H.G.Muller wrote:This option would be much more logical on using the book. Then you can keep the information on the move weights in the book, but simply ignore it when picking a move.


... but you have no way to get the precise choice of moves I get in my little example (forced e4 or d4 then moves according to PG heuristics on a large database).

In fact the precise goal of adding "book-merge", "only-white/only-black" and "book-merge" to the book making features of Polyglot was the possibility to combine fixed-lines with the broader knowledge coming from large databases in view of preparing Fruit competition books at summer 2005.
Don't forget that we had no way to change the weights "manually" once the book had been built at the time (this came much later with P George's added features in SCID)

This mechanism was elaborated by Fabien after quite a series of mail exchanges with myself.

Marc
Marc Lacrosse
 
Posts: 116
Joined: 29 Jan 2005, 09:04
Location: Belgium


Return to Winboard and related Topics

Who is online

Users browsing this forum: No registered users and 31 guests