Posts : 1254 Join date : 2020-11-17 Location : France
Subject: Say no to opening books Mon Apr 18, 2022 12:12 am
The game between the two second equals. We really should get rid of these books. NN based engines can perfectly well handle openings without needing to look the moves up in a book. This one was 40 moves long, it's ridiculous. Trouble is, because everybody does it, you have to as well. I used to use a very short book, and also played without any book at all, with the result of running into the same prepared line of Marshall Gambit game after game against several different opponents. I guess they all use the same book. It's not very interesting to compete this way.
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: Say no to opening books Mon Apr 18, 2022 12:27 am
Or a book that throws the others OUT of their nonsense big books.
adminx likes this post
TheSelfImprover
Posts : 3112 Join date : 2020-11-18
Subject: Re: Say no to opening books Mon Apr 18, 2022 2:39 pm
Opening book: noun: a database of chess moves for the early phase of the game useful in chess programs from around 1970 to about 2020
adminx likes this post
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: Say no to opening books Mon Apr 18, 2022 2:50 pm
I repeat my thesis: sharp gambit opening lines for almost any known opening. Then throwing other OUT of book by strange minor moves such as 1.a3, h3 or 1…a6 or h6.
Peter Berger
Posts : 131 Join date : 2020-11-20
Subject: Re: Say no to opening books Mon Apr 18, 2022 7:50 pm
Chris Whittington wrote:
The game between the two second equals. We really should get rid of these books. NN based engines can perfectly well handle openings without needing to look the moves up in a book. This one was 40 moves long, it's ridiculous. Trouble is, because everybody does it, you have to as well. I used to use a very short book, and also played without any book at all, with the result of running into the same prepared line of Marshall Gambit game after game against several different opponents. I guess they all use the same book. It's not very interesting to compete this way.
I do most of the games I run with one of the engines ( usually the stronger one) using no opening book. I agree that modern engines do very well without a book in general, especially at slower time controls. But there are still a few lines where you can catch them with the accumulated knowledge of humans aided by the help of engines. These lines are few though, and in all automatically generated books there is probably way more garbage than value percentage-wise. I think you can do very well with a small book, but it needs some work, and it is not the kind of work programmers tend to like, as the core of it can’t really be fully automated IMHO without producing similar holes. But the amount of work that is needed for a decent book isn’t a lot – one hundred working hours should be enough for something that cannot be caught. What you need is something like the book.bin of crafty that eats PGN and can deal with ! and ?, so that you can force and avoid some moves. In most of the cases the engine can just play by itself. This is how I used to do books myself, and I was always amazed by the fact how little time is really needed to come up with sth decent. What Marshall Gambit are we talking about here btw? If you use the move order 1. d4 d5 2. d4 e6 3. Nc3 c6 4. e4 you avoid the Slav exchange that I consider to be very difficult for computers ( while the ordinary queens gambit exchange line isn’t that much of a threat to modern engines anymore), and I think there are very safe and reliable lines against 4. e4. You can probably deal with this setup by forcing like ten moves in total. If you are talking about the Ruy Lopez Marshall gambit and someone caught you in a drawn line ( you being white), well the Marshall is just a draw these days . But if you go 8. a4 you‘ll still get a full game. I don’t think there is any real use in lines that are 20 moves long. It’s just that most programmers do their books in similar ways and decide to let the books get that deep. Peter
Uri Blass
Posts : 207 Join date : 2020-11-28
Subject: Re: Say no to opening books Tue Apr 19, 2022 4:11 pm
Peter Berger wrote:
Chris Whittington wrote:
The game between the two second equals. We really should get rid of these books. NN based engines can perfectly well handle openings without needing to look the moves up in a book. This one was 40 moves long, it's ridiculous. Trouble is, because everybody does it, you have to as well. I used to use a very short book, and also played without any book at all, with the result of running into the same prepared line of Marshall Gambit game after game against several different opponents. I guess they all use the same book. It's not very interesting to compete this way.
I do most of the games I run with one of the engines ( usually the stronger one) using no opening book. I agree that modern engines do very well without a book in general, especially at slower time controls. But there are still a few lines where you can catch them with the accumulated knowledge of humans aided by the help of engines. These lines are few though, and in all automatically generated books there is probably way more garbage than value percentage-wise. I think you can do very well with a small book, but it needs some work, and it is not the kind of work programmers tend to like, as the core of it can’t really be fully automated IMHO without producing similar holes. But the amount of work that is needed for a decent book isn’t a lot – one hundred working hours should be enough for something that cannot be caught. What you need is something like the book.bin of crafty that eats PGN and can deal with ! and ?, so that you can force and avoid some moves. In most of the cases the engine can just play by itself. This is how I used to do books myself, and I was always amazed by the fact how little time is really needed to come up with sth decent. What Marshall Gambit are we talking about here btw? If you use the move order 1. d4 d5 2. d4 e6 3. Nc3 c6 4. e4 you avoid the Slav exchange that I consider to be very difficult for computers ( while the ordinary queens gambit exchange line isn’t that much of a threat to modern engines anymore), and I think there are very safe and reliable lines against 4. e4. You can probably deal with this setup by forcing like ten moves in total. If you are talking about the Ruy Lopez Marshall gambit and someone caught you in a drawn line ( you being white), well the Marshall is just a draw these days . But if you go 8. a4 you‘ll still get a full game. I don’t think there is any real use in lines that are 20 moves long. It’s just that most programmers do their books in similar ways and decide to let the books get that deep. Peter
What is the target of a book? Beating stockfish15 with no book at long time control or maybe beating strong engines that are weaker than stockfish.
I think that it may be interesting if people produce a book to show how to beat stockfish15 with no book at long time control if they can do it (it is better with a single core so games may be reproducable).
Ray
Posts : 38 Join date : 2020-11-26
Subject: Re: Say no to opening books Thu May 05, 2022 9:10 pm
To me, Chess960 is the obvious solution to this opening book "problem". It is why I am such a fan of it, and I play very few standard chess games these days.
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: Say no to opening books Wed May 11, 2022 11:21 pm
Home 
