Value of a really good book

[Dann Corbit]

Polhl's SF testing page shows that a good book ( Brainfish's book called Cerebellum ) is worth about 35 Elo or so.
I wonder how useing Bojun Guo's mighty database would fare.
Is there a version of Stockfish set up to use his database as a book using his API directly?

[Ovyron]

Brainfish's Cerebellum is nowhere close to being a "really good book", get a free playchess account to see how its lines have been busted and get a beating by the good books people built themselves.

[jdart]

Sedat Canbaz has run a series of book tournament (same engine, different books). The results give some indiciation of how much a book matters, and also indicate which of the books he is using perform the best:

https://sites.google.com/site/computerschess/home

[yurikvelo]

Polhl's SF testing page shows that a good book ( Brainfish's book called Cerebellum ) is worth about 35 Elo or so.

under bullet STC

[Dann Corbit]

Brainfish's Cerebellum is nowhere close to being a "really good book", get a free playchess account to see how its lines have been busted and get a beating by the good books people built themselves.

It gets rebuilt all the time.
It recently won one of Sedat's contests, after getting killed for a while.

[Dann Corbit]

Brainfish's Cerebellum is nowhere close to being a "really good book", get a free playchess account to see how its lines have been busted and get a beating by the good books people built themselves.

It gets rebuilt all the time. I would not be surprised if they used Bojun's data.
It recently won one of Sedat's contests, after getting killed for a while.

See, for example, this recent contest:
https://sites.google.com/site/computers ... book-cs-42

[mehmet123]

This tournament was a rebuilt (by Sedat Canbaz) opening book tournament.
Don (by Vasid Chouhan)book is the strongest opening book.It had won nearly all the tournaments it attended in 2020

[mehmet123]

These were the original opening book tournaments.

https://sites.google.com/site/computers ... -cs-40-pro

https://sites.google.com/site/computers ... book-cs-36

Don Opening Book vs No opening book difference was +137 elo

[mehmet123]

DonLite2 Opening Book vs Deep Fritz 10 Opening Book Match.

Deep Fritz 10 is the opening book of the engine that beat World Champion Kramnik with a 4-2 score at 2006.
https://en.chessbase.com/post/kramnik-v ... tch-by-4-2

I did this test to see the level that opening books have reached in 14 years. The difference is really huge.

Program Elo + - Games Score Av.Op. Draws

1 Stockfish 120520 64 BMI2 (DonLite2) : 2512 41 36 100 65.5 % 2400 65.0 %
2 Stockfish 120520 64 BMI2 x1 (Deep Fritz 10): 2400 36 41 100 34.5 % 2512 65.0 %

Individual statistics:

1 Stockfish 120520 64 BMI2 : 2512 100 (+ 33,= 65,- 2), 65.5 %

Stockfish 120520 64 BMI2 x1 : 100 (+ 33,= 65,- 2), 65.5 %

2 Stockfish 120520 64 BMI2 x1: 2400 100 (+ 2,= 65,- 33), 34.5 %

Stockfish 120520 64 BMI2 : 100 (+ 2,= 65,- 33), 34.5 %

Fritz 17 Gui, 3 minutes tc, 1 core, ponder off
Opening books set to recommended settings.
http://www.mediafire.com/file/4scwdtxdv ... 1.pgn/file

[Ovyron]

See, for example, this recent contest

Those are laboratory conditions. As I said, you can go to playchess and see Cerebellum being destroyed to pieces by the books of people there (books Sedat does not have.) I believe Wilson Fisk has the strongest book, today he made it to 2800 elo there. Cevdet will get destroyed. Don will get destroyed. You need daily work on your private book to avoid destruction.

To answer your original question, by my estimate at least 100 elo can be added on top of Cerebellum, so the value of a really good book would be around +237 elo (over no book.)

(this is at 5 +0 time control average, YMMV)