Page 4 of 10

Re: CCRL 40/40 - A most curious game

Posted: Mon Oct 26, 2009 12:52 pm
by Mike S.
pichy wrote:I will not post them all since it will take too much space on CCC.
Most of the space is wasted with mainline for every move. In Arena, press [F7] --> PGN and there you can switch that off.

(Eval+depth+time is a different options and can remain.)

Also, I don't see what sense it makes to simply post games. If you want to support your idea, you should at least point to the particular opening moves you mean, with some comments and analysis. Also, one game for one move won't tell much...

Re: CCRL 40/40 - A most curious game

Posted: Mon Oct 26, 2009 5:48 pm
by pichy
Mike S. wrote:
pichy wrote:I will not post them all since it will take too much space on CCC.
Most of the space is wasted with mainline for every move. In Arena, press [F7] --> PGN and there you can switch that off.

(Eval+depth+time is a different options and can remain.)

Also, I don't see what sense it makes to simply post games. If you want to support your idea, you should at least point to the particular opening moves you mean, with some comments and analysis. Also, one game for one move won't tell much...
The only reason for posting some of those games, was to show that even if the opening are limited and most of the times Rybka select the same 1.Nb3 first move for most of its games it still can hold against some very fine but old Arena.abk opening book.

PS: But for those looking for a variety of opening lines a good opening book like your and Dr. Wael Deeb will be more than enough.

Re: CCRL 40/40 - A most curious game

Posted: Mon Oct 26, 2009 6:14 pm
by Guenther
pichy wrote:
Mike S. wrote:
pichy wrote:I will not post them all since it will take too much space on CCC.
Most of the space is wasted with mainline for every move. In Arena, press [F7] --> PGN and there you can switch that off.

(Eval+depth+time is a different options and can remain.)

Also, I don't see what sense it makes to simply post games. If you want to support your idea, you should at least point to the particular opening moves you mean, with some comments and analysis. Also, one game for one move won't tell much...
The only reason for posting some of those games, was to show that even if the opening are limited and most of the times Rybka select the same 1.Nb3 first move for most of its games it still can hold against some very fine but old Arena.abk opening book.

PS: But for those looking for a variety of opening lines a good opening book like your and Dr. Wael Deeb will be more than enough.
Uri was right and your test is flawed (again).
Even from the small sample you showed one can see that
you did not test what you wanted, because _both_ programs
didn't use a book each second game.
(Also the few games when it seems one program used a book
the openings are surprisingly without much variation?)
IMO it is also not very good to use short suddendeath time controls
in a book test, because equal endgames become a lottery with 0 sec
moves and the book/no book setting has nothing to do with the result then.

Guenther

Re: CCRL 40/40 - A most curious game

Posted: Mon Oct 26, 2009 11:47 pm
by pichy
Guenther wrote:
pichy wrote:
Mike S. wrote:
pichy wrote:I will not post them all since it will take too much space on CCC.
Most of the space is wasted with mainline for every move. In Arena, press [F7] --> PGN and there you can switch that off.

(Eval+depth+time is a different options and can remain.)

Also, I don't see what sense it makes to simply post games. If you want to support your idea, you should at least point to the particular opening moves you mean, with some comments and analysis. Also, one game for one move won't tell much...
The only reason for posting some of those games, was to show that even if the opening are limited and most of the times Rybka select the same 1.Nb3 first move for most of its games it still can hold against some very fine but old Arena.abk opening book.

PS: But for those looking for a variety of opening lines a good opening book like your and Dr. Wael Deeb will be more than enough.
Uri was right and your test is flawed.
Even from the small sample you showed one can see that
you did not test what you wanted, because _both_ programs
didn't use a book each second game.
(Also the few games when it seems one program used a book
the openings are surprisingly without much variation?)
IMO it is also not very good to use short suddendeath time controls
in a book test, because equal endgames become a lottery with 0 sec moves and the book/no book setting has nothing to do with the result then.

Guenther
I tested exactly what I wanted, since the second program which is Rybka 2.32a-B without an Opening Book was meant to be tested versus the exact Rybka 2.32a-A which is using Rybka.abk. The reason for coming up with this experiment is to justify the argument that if you are going to use an opening book which is not filted agains errors, in that case it is much better to simply not use any opening book and the program will perform better, like in the case of Bright vs Cyclone xTreme. As for the time control of 5 minutes it might not be enough time specially for the program not using an opening book, since a lot of time is consumed in searching for the next best move from the start of the game. Therefore, more time will simply benefit the handicapped program, not the other way around. Now in time control of less than 5 minutes than you have a serious problem, in which you are not allowing the handicapped program enough time to search for the best next available move, and eventually it will run out off time for searching fot the best move during the endgame stage.

Re: CCRL 40/40 - A most curious game

Posted: Tue Oct 27, 2009 3:19 am
by tano-urayoan
Following the games you posted there is a problem. In the first and third neither engine is using a book. In the second and fourth one the book was just 2 moves.

You should check your configuration for this test.

Re: CCRL 40/40 - A most curious game

Posted: Tue Oct 27, 2009 11:31 am
by pichy
tano-urayoan wrote:Following the games you posted there is a problem. In the first and third neither engine is using a book. In the second and fourth one the book was just 2 moves.

You should check your configuration for this test.
I noticed that some games for some unknown reasons the book was just two or three moves. But I do have the settings correctly.

On the left side of the move selection under book Manage where it say minimal conditions to play the move.

1. minimal number of games = 2

------------------------------------------------------------------------------------
On the right hand side under probability set by......

1.Use the opening book until the 32 halfmoves.
2.Set Priority +15,Number of games +15,Win percentage 0

Re: CCRL 40/40 - A most curious game

Posted: Tue Oct 27, 2009 12:29 pm
by Dr.Wael Deeb
tano-urayoan wrote:Following the games you posted there is a problem. In the first and third neither engine is using a book. In the second and fourth one the book was just 2 moves.

You should check your configuration for this test.
He's using the right configuration but unfortunately his test shows no sense as every time Rybka plays 1.Nc3,there is not much to be answered with....
Dr.D

Re: CCRL 40/40 - A most curious game

Posted: Tue Oct 27, 2009 5:12 pm
by Guenther
Dr.Wael Deeb wrote:
tano-urayoan wrote:Following the games you posted there is a problem. In the first and third neither engine is using a book. In the second and fourth one the book was just 2 moves.

You should check your configuration for this test.
He's using the right configuration but unfortunately his test shows no sense as every time Rybka plays 1.Nc3,there is not much to be answered with....
Dr.D
Do you mean the used book has _zero_ book moves against 1. Nc3?
Well he could have noticed this already in the second game then ;-)

Guenther

Re: CCRL 40/40 - A most curious game

Posted: Tue Oct 27, 2009 5:18 pm
by Guenther
pichy wrote:
Guenther wrote:
pichy wrote:
Mike S. wrote:
pichy wrote:I will not post them all since it will take too much space on CCC.
Most of the space is wasted with mainline for every move. In Arena, press [F7] --> PGN and there you can switch that off.

(Eval+depth+time is a different options and can remain.)

Also, I don't see what sense it makes to simply post games. If you want to support your idea, you should at least point to the particular opening moves you mean, with some comments and analysis. Also, one game for one move won't tell much...
The only reason for posting some of those games, was to show that even if the opening are limited and most of the times Rybka select the same 1.Nb3 first move for most of its games it still can hold against some very fine but old Arena.abk opening book.

PS: But for those looking for a variety of opening lines a good opening book like your and Dr. Wael Deeb will be more than enough.
Uri was right and your test is flawed.
Even from the small sample you showed one can see that
you did not test what you wanted, because _both_ programs
didn't use a book each second game.
(Also the few games when it seems one program used a book
the openings are surprisingly without much variation?)
IMO it is also not very good to use short suddendeath time controls
in a book test, because equal endgames become a lottery with 0 sec moves and the book/no book setting has nothing to do with the result then.

Guenther
I tested exactly what I wanted, since the second program which is Rybka 2.32a-B without an Opening Book was meant to be tested versus the exact Rybka 2.32a-A which is using Rybka.abk. The reason for coming up with this experiment is to justify the argument that if you are going to use an opening book which is not filted agains errors, in that case it is much better to simply not use any opening book and the program will perform better, like in the case of Bright vs Cyclone xTreme. As for the time control of 5 minutes it might not be enough time specially for the program not using an opening book, since a lot of time is consumed in searching for the next best move from the start of the game. Therefore, more time will simply benefit the handicapped program, not the other way around. Now in time control of less than 5 minutes than you have a serious problem, in which you are not allowing the handicapped program enough time to search for the best next available move, and eventually it will run out off time for searching fot the best move during the endgame stage.
You completely missed the point of my post. Didn't you notice that
each second game was played without _any_ book moves.
(I wonder why I must repeat what I already had written before.
It would be nice if you would read my post before you answer
unrelated stuff and just take a look at your posted games, sigh.)

Code: Select all

Example:

1. Nc3 {(Nb1c3 Nb8c6 Ng1f3 Ng8f6 d2d4 d7d5 Bc1f4 e7e6 e2e3 Bf8d6) +0.13/13
9} Nf6 {(Ng8f6 Ng1f3 Nb8c6 e2e4 d7d5 e4xd5 Nf6xd5 Bf1c4 Bc8e6 Bc4xd5 Be6xd5
OO) -0.07/12 10}
There is not much to missunderstand neither in my post nor
in the pgn quote above.

Guenther

Re: CCRL 40/40 - A most curious game

Posted: Tue Oct 27, 2009 5:45 pm
by tano-urayoan
pichy wrote:
I noticed that some games for some unknown reasons the book was just two or three moves. But I do have the settings correctly.

On the left side of the move selection under book Manage where it say minimal conditions to play the move.

1. minimal number of games = 2

------------------------------------------------------------------------------------
On the right hand side under probability set by......

1.Use the opening book until the 32 halfmoves.
2.Set Priority +15,Number of games +15,Win percentage 0
Again as Gunther Simon explains in the games you posted games 1 and 3 neither of the engines used a book. And in the other 2 games the book was just 2 moves.

The test you are conducting is biased because really no book is being used by neither of the engines.