ChessKnight wrote:Hi Marc Lacrosse
According to your suggestion, i have started organizing PGN files.
I have following questions for you.
1) The pgn for draw games should be excluded?
I do not think so. In fact, discarding drawn games may have a paradoxical consequence : the book itself will be drawish!
See here :
http://rybkaforum.net/cgi-bin/rybkaforu ... 100608;hl=ChessKnight wrote:2) in the code given by you
The most simple way is to do something like :
Code:
polyglot.exe make-book -pgn MyGameCollection.pgn -max-ply 50 -min-game 3 -min-score 30 -bin MyPolyglotBook.bin
Can you explain min-game and min-score parameters?
I repeat my recommendation : do read polyglot's manual !
-min-game 3 : in a given position, only the moves that have been played in at least three games will be included in the book. A move that has been played only in one game will not be included in the book even though this precise game has been won by the side to move.
-min-score : similar to the min-game parameter, but deals with global results : only thoses moves that have achieved an overall result superior or equal to your given value for the side considered (in percents) will be included in the book.
ChessKnight wrote:What are the best settings for above command?
It depends on what you are trying to achieve with your book.
Some people have worked for years in book making and tuning ...
ChessKnight wrote:(...)What is the exact procedure for tuning an opening book?
ChessKnight wrote:4) Is there any easiest way to check if opening book created is enough good?
Same answer as former one.
Basically do have your engine(s) play lots of games with your book and do try to correct under-performing lines. There is no published automated way to achieve this, so it is still a matter of knowledge, feeling and experience supported by many many hours of "manual" analysis mixed with many many hours of testing.
This is a point where the better chess player has still an enormous advantage over a weak one. But there are well-known grandmasters who are credited with engine opening books of rather mediocre quality.
ChessKnight wrote:5) When a fresh newly made opening is linked with any chess engine is there certain amount of games needs to played to auto-tune opening book?
Polyglot books do not auto-tune.
The more the games the more you will have stuff for improving the book.
ChessKnight wrote:6) Up to how many initial moves normally a chess engine depends on an opening book?
This is very controversial.
If your goal is to have better results in typical online engine play (blitz on Playchess ...), the simple answer is "longer is better", because as long as your engine is in his book, it does not consume any time and will get more time kept for when it will begin to think on its own.
But if you push it to the extreme your engine won't play at all.
So if you are comfident that your engine is superior to his opponents in early middle game you should not make a too long book. Recent posts by Jeroen Noomen and Vas Rajlich explain how on this precise ground they decided to go for shorter competition books for Rybka avoiding overanalysed main opening variations ...
If your goal is not "results, better results, always better results" in a very limited set of conditions (playing blitz with rybka 3 against other rybka 3 on Playchess) then there is no possible general answer.
You could also have a look at this thread here :
http://www.open-aurec.com/wbforum/viewt ... highlight=ChessKnight wrote:Thanking you for your time and assitance.
You are welcome
Marc