Good books for self testing?

[CRoberson]

I am trying perfekt 2010. However, saw the early stats very different from the later stats.
While this can just be a stats issue it could be a book issue. Any suggestions for better books.

Charles

[xr_a_y]

I always use AH books, from here : http://kirill-kryukov.com/chess/tools/o ... mpler/pgn/

[jp]

I am trying perfekt 2010.

How many opening positions does it have? Oh, do you mean perfect 2010? (I cannot find perfekt 2010 online.)

I always use AH books, from here : http://kirill-kryukov.com/chess/tools/o ... mpler/pgn/

Are those numbers at the end of the filenames the number of openings contained?

[xr_a_y]

It seems.

[jp]

Maybe helpful:

http://talkchess.com/forum3/viewtopic.php?f=2&t=67502
linking to
http://rebel13.nl/download/utilities.html#one

If anyone tries this, tell us what you think.

[Albert Silver]

Here is a large suite, all handpicked, and all real openings, with none of that random 3-move nonsense. Every opening under the sun.

[jp]

Albert, handpicked meaning by you? How did you approach that?
And how deep is each opening roughly?

[lucasart]

Here is a large suite, all handpicked, and all real openings, with none of that random 3-move nonsense. Every opening under the sun.

It's obvious your not an engineer. The "random 3-move nonsense" is exactly what chess engine programmers need for self testing based development. If you double the elo resolution of the book, you divide by 4 the number of games required to reach any given precision. And testing is (by far) the bottleneck of engine development.

[Albert Silver]

Here is a large suite, all handpicked, and all real openings, with none of that random 3-move nonsense. Every opening under the sun.

It's obvious your not an engineer. The "random 3-move nonsense" is exactly what chess engine programmers need for self testing based development. If you double the elo resolution of the book, you divide by 4 the number of games required to reach any given precision. And testing is (by far) the bottleneck of engine development.

It is one way of looking at it.

The other is that the openings all represent distinct characteristics, with varied structures, and subsequent endgames. If you leave it to 3-move books, you may certainly see if your engine is able to consistently reach a position it favors, but will have no idea of the holes in its understanding or play, because they never appeared, much less plug them.

[Albert Silver]

Albert, handpicked meaning by you? How did you approach that?
And how deep is each opening roughly?

Remember I am a chess player, so I do come with knowledge of the openings, and what exists. I tried to be logical, covering the entire gamut of openings, with no bias or preference on my end. The proportion of the openings is also close to the frequency in the Mega Database, so clearly some openings will be more represented than others, but when some openings would be underrepresented due to lack of play, I gave them a nudge to make sure they had a few.

The depth is complicated since some openings are far more explored than others. I do not mean I am pushing the boundaries of theory, since this is an openings suite not an openings book. For example, the Benko Gambit is defined by 3...b5, after which there are many lines, accepted and declined. They are there to see how the engine handles each and every one. Bird, Italian, Spanish, KID, KIA, you name it, it is there, but just enough moves to define each line, and the moves before that could lead to it.

For me a chess engine needs to show proficiency in all types of positions, and this suite aims to help achieve that aim.