Random neutral .abk book for Arena?

[Mike S.]

It exists! Although, not intended as a "performance" book at the highest levels, but as a book with more variety (wider repertoire), for engine testing with same book for all. I have made several books of that type: Xmas2640 for Fritz, Arena and Shredder interfaces, and Balanced-12 and Balanced-16 for Fritz interfaces (the number is max. lenght in full moves).

http://members.aon.at/computerschach/li ... #downloads

These books contain no engine specific tuning and not even any computer chess related tuning. My opinion is that engines should play everything what their users might want to see, not what the engines like themselves.

You can see how (surprisingly good) these books rank in S.Canbaz' book test tournament:

http://www.sedatchess.com/scct_opening_book.html

(see the 2nd table below, for the Arena books)

Of course, these results cannot be seen independant from the engines used, playing against themselves with different books: mostly Rybka, and Zappa. I cannot estimate if my (or anyone else's) books would rank better or worse, if other engines would be used.

[bob]

I don't think such a book exists. If the book includes an opening some specific engine does not play very well, even though the positions reached are perfectly equal, that program will be at a disadvantage since it is being forced to play an opening it doesn't understand...

Hi Bob,

I am not a programmer, but I would have thought that engines would be programmed to understand general opening principles rather than specific openings?
General opening principles would apply to most openings?

Regards, Graham.

It really does depend on a lot of things. While you are correct on a general plane, in reality, openings have "themes" and some programs excel at some themes and play poorly against others. For example, in some openings, a kingside attack is the issue, and many programs are clueless there. So avoiding those kinds of openings makes that particular program play stronger because it avoids the openings it doesn't understand as well. I know many humans that play the same way. Some hate complicated positions, some play for them intentionally... More subtle issues are also at play, such as endgame pawn structure, or a weak/isolated queen's pawn, etc... I have seen programs avoid opposite-castle openings for similar reasons... There is significant strength to be had by avoiding what you don't play very well... until the holes are fixed...

[bob]

I don't think such a book exists. If the book includes an opening some specific engine does not play very well, even though the positions reached are perfectly equal, that program will be at a disadvantage since it is being forced to play an opening it doesn't understand...

Hi Bob,

I am not a programmer, but I would have thought that engines would be programmed to understand general opening principles rather than specific openings?
General opening principles would apply to most openings?

Regards, Graham.

It really does depend on a lot of things. While you are correct on a general plane, in reality, openings have "themes" and some programs excel at some themes and play poorly against others. For example, in some openings, a kingside attack is the issue, and many programs are clueless there. So avoiding those kinds of openings makes that particular program play stronger because it avoids the openings it doesn't understand as well. I know many humans that play the same way. Some hate complicated positions, some play for them intentionally... More subtle issues are also at play, such as endgame pawn structure, or a weak/isolated queen's pawn, etc... I have seen programs avoid opposite-castle openings for similar reasons... There is significant strength to be had by avoiding what you don't play very well... until the holes are fixed...

[Graham Banks]

I don't think such a book exists. If the book includes an opening some specific engine does not play very well, even though the positions reached are perfectly equal, that program will be at a disadvantage since it is being forced to play an opening it doesn't understand...

Hi Bob,

I am not a programmer, but I would have thought that engines would be programmed to understand general opening principles rather than specific openings?
General opening principles would apply to most openings?

Regards, Graham.

It really does depend on a lot of things. While you are correct on a general plane, in reality, openings have "themes" and some programs excel at some themes and play poorly against others. For example, in some openings, a kingside attack is the issue, and many programs are clueless there. So avoiding those kinds of openings makes that particular program play stronger because it avoids the openings it doesn't understand as well. I know many humans that play the same way. Some hate complicated positions, some play for them intentionally... More subtle issues are also at play, such as endgame pawn structure, or a weak/isolated queen's pawn, etc... I have seen programs avoid opposite-castle openings for similar reasons... There is significant strength to be had by avoiding what you don't play very well... until the holes are fixed...

Thanks Bob,

your explanation makes perfect sense.

Regards, Graham.