I appreciate if vas or a rybka team member replies.
Thanks in advance.
- at least +100 elo average score against the latest public books in the initial release, (supposing rybka 4 is +100 then rybka 4 book should be also +100 against public books).
- It should not score below 40% against any public book in 100 games.
- Has updates at least twice a month or weekly. And a description of which lines were edited in the update.
All book testing should be done with same engine, and not giving the public book with a weaker one of course.
How a book performs in Sedat's book tests is meaningless to the vast majority of chess players, most of whom are looking for a well rounded book to practice/prepare with or one that can be used as a general reference. The Playchess engine room crowd is a very small segment of the market and as you mentioned, it's a segment that's served very well by the seemingly endless stream of free public books. Most of these free books seem to completely ignore 75% of the ECO which makes them rather useless as a reference point or a practice tool. Width matters.
Program Elo + - Games Score Av.Op. Draws
1 Pferdchen 3 : 3357 28 28 340 65.9 % 3243 43.5 %
2 My book 17.1 : 3349 43 42 130 60.4 % 3275 50.0 %
3 My book 17 : 3308 28 28 290 58.8 % 3246 52.1 %
4 My book 16 : 3289 19 19 620 58.1 % 3232 53.2 %
....66 Rybka 3 Book : 3137 32 33 190 45.3 % 3170 56.8 %
And only 20 elo stronger vs nearest rival on initial release: http://www.sedatchess.com/scct_book.html
And they have monster hardware and rybka beta in making it, I have no beta versions and my 'pferdchen 3' got +50 elo vs nearest rival on initial release and I made it alone with only 1 average quad core pc. And best it is free =)
>Rybka 3 book now is one of the weakest book for computer chess.
Total nonsense. Make 2 little changes in that book to avoid Najdorf (one for white, one for black) and it will beat most other books. I used that book in several cluster vs. centaur games and it worked well.
>Make 2 little changes in that book to avoid Najdorf
First you don't change it because that's what the author's intention of marking the lines. If you change it, that's not rybka 3 book anymore.
>I used that book in several cluster vs. centaur games and it worked well.
No, that's not a proper way to test book strength (uneven hardware and engine). Plus with little games it is more like a statistical noise.
After a while that arms race just got silly/boring, for me.
>Make 2 little changes in that book to avoid Najdorf (one for white, one for black) and it will beat most other books.
You can of course send that rybka 3 book with 2 changes to sedat and test it versus a batch of public books made in 2009 with even hardware and engine, a good and simple way of proving your theory. My bet is that it won't even hit top 5, maybe $1000 on stake and I want your rybka cluster software plus $500 if you lose :-)
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill