KUT-O2 / game 4: Crafty1711 vs Gromit30 ( 0-1)

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 03 July 2000 15:23:22:
Als Antwort auf: / As an answer to: Re: About openibg book creation geschrieben von: / posted by: Aaron at 03 July 2000 15:08:43:

BTW i'm looking at your book you created for Yace..For a moment, it seemed that Yace as white played 1 Nf3 almost excusively, and sicilian as black..
At first, I did a few manual games to test out the lines. But this was very time consuming, since after every game I had to restart Winboard.
Apparantly, after resigning and then resetting the board, yace has problems restarting..
Perhaps I should try manually, the Amy book on your site..They are based on almost the same games right?? Of course I understand there are some differences in book creation..

Really? I wasn't aware of that.
Yes, almost the same games. Positions that only appear once where deleted when I used the "flatten 1" command. Shouldn't make an overwhelming difference.
Best wishes...
Mogens

[Aaron]

Geschrieben von: / Posted by: Aaron at 03 July 2000 15:58:23:
Als Antwort auf: / As an answer to: Re: About openibg book creation geschrieben von: / posted by: Mogens Larsen at 03 July 2000 15:23:22:

BTW i'm looking at your book you created for Yace..For a moment, it seemed that Yace as white played 1 Nf3 almost excusively, and sicilian as black..
At first, I did a few manual games to test out the lines. But this was very time consuming, since after every game I had to restart Winboard.
Apparantly, after resigning and then resetting the board, yace has problems restarting..
Perhaps I should try manually, the Amy book on your site..They are based on almost the same games right?? Of course I understand there are some differences in book creation..

Really? I wasn't aware of that.
Yes, almost the same games. Positions that only appear once where deleted when I used the "flatten 1" command. Shouldn't make an overwhelming difference.
Best wishes...
Mogens

[Aaron]

Geschrieben von: / Posted by: Aaron at 03 July 2000 20:20:46:
Als Antwort auf: / As an answer to: Opening book creation for Yace geschrieben von: / posted by: Dieter Buerssner at 03 July 2000 14:21:30:
Thanks for all your help..It works!!!..But I have more questions..

!copy /Y 2600.pgn+aaron.pgn book.pgn
Finally I understand this line..This creates a combined Book.pgn which is later converted into the new opening book..

How exactly do you know, you have a good storage size?
I got
nused 97975
0: 2025
1: 79110
2: 11488
3: 4106

etc etc. Most efficent size occurs when "nused" is as close to zero as possible or is
0: X , where X is as small as zero as possible best?
Sorry, I read your help file and am still a little confused..

[Dieter Buerssner]

Geschrieben von: / Posted by: Dieter Buerssner at 04 July 2000 17:59:46:
Als Antwort auf: / As an answer to: Re: Opening book creation for Yace geschrieben von: / posted by: Aaron at 03 July 2000 20:20:46:

How exactly do you know, you have a good storage size?
I got
nused 97975
0: 2025
etc etc. Most efficent size occurs when "nused" is as close to zero as possible or is

This looks fine. If the 0: line is between about 1 and 30%, everything is
ok. Values smaller one percent, will have worse performance, values > 30%
will waste much disk space. nused cannot get zero, because this is the number
of book positions.
Regards,
Dieter

[Aaron]

Geschrieben von: / Posted by: Aaron at 04 July 2000 18:26:28:
Als Antwort auf: / As an answer to: Re: Opening book creation for Yace geschrieben von: / posted by: Dieter Buerssner at 04 July 2000 17:59:46:

How exactly do you know, you have a good storage size?
I got
nused 97975
0: 2025
etc etc. Most efficent size occurs when "nused" is as close to zero as >>possible or is

This looks fine. If the 0: line is between about 1 and 30%, everything is
ok. Values smaller one percent, will have worse performance, values 30%
will waste much disk space. nused cannot get zero, because this is the number
of book positions.

Forgive me for asking again..
1% to 30% of what figure?

I've also being thinking about the use of ! or ? for marking Yace opening books.
Do you have to mark the PGN files with such comments then complie it? That would be quite troublesome for big books..
I would sugguest something like the Amy format, where there is a opening book and a "learn" or "preference file"
Thank you for your answers..

Aaron

[Dieter Buerssner]

Geschrieben von: / Posted by: Dieter Buerssner at 04 July 2000 20:15:59:
Als Antwort auf: / As an answer to: Re: Opening book creation for Yace geschrieben von: / posted by: Aaron at 04 July 2000 18:26:28:

How exactly do you know, you have a good storage size?
I got
nused 97975
0: 2025
etc etc. Most efficent size occurs when "nused" is as close to zero as >>>possible or is

This looks fine. If the 0: line is between about 1 and 30%, everything
is ok.

1% to 30% of what figure?
I've also being thinking about the use of ! or ? for marking Yace
opening books.
Do you have to mark the PGN files with such comments then complie it?
That would be quite troublesome for big books..

The number, you have given as the initial size. After rethinking,
I assume, that a good value for the initial size should be the
number shown as nused. No experience with large books though.
I assume the numbers you have posted here are not about the
large book from the other message. nused is the number of different
book positions, it would seem much too low for such a large book.
How long did the creation of the large book take?
Also, please note, that all public available versions of Yace have
a bug in the book move selection code, when the sum of counts for the
different moves (Yace shows them in the log) is very high. Then, it
will just not consider moves further down in the list. I think, there
should be no problem with the book created from 2600.pgn, but I
suspect, that with your large book, you will see it. If you like,
I can mail you a fixed version.
I first thought about a scheme, like the one I suggested to you
for adding book lines. Just copy the original PGN file and your
marked lines together and create a new book.
It might however be more convenient for users, to have a crafty like
books.bin. I will think about it.
Book learning will need some time ...
Regards,
Dieter

[Aaron]

Geschrieben von: / Posted by: Aaron at 04 July 2000 21:50:36:
Als Antwort auf: / As an answer to: Re: Opening book creation for Yace geschrieben von: / posted by: Dieter Buerssner at 04 July 2000 20:15:59:

How exactly do you know, you have a good storage size?
I got
nused 97975
0: 2025
etc etc. Most efficent size occurs when "nused" is as close to zero as >>>possible or is This looks fine. If the 0: line is between about 1 and 30%, >>>everything is ok

I've also being thinking about the use of ! or ? for marking Yace
opening books.
Do you have to mark the PGN files with such comments then complie it?
That would be quite troublesome for big books..

1% to 30% of what figure?
The number, you have given as the initial size. After rethinking,
I assume, that a good value for the initial size should be the
number shown as nused. No experience with large books though.
I assume the numbers you have posted here are not about the
large book from the other message. nused is the number of different
book positions, it would seem much too low for such a large book.
How long did the creation of the large book take?
Also, please note, that all public available versions of Yace have
a bug in the book move selection code, when the sum of counts for the
different moves (Yace shows them in the log) is very high. Then, it
will just not consider moves further down in the list. I think, there
should be no problem with the book created from 2600.pgn, but I
suspect, that with your large book, you will see it. If you like,
I can mail you a fixed version.
I first thought about a scheme, like the one I suggested to you
for adding book lines. Just copy the original PGN file and your
marked lines together and create a new book.
It might however be more convenient for users, to have a crafty like
books.bin. I will think about it.
Book learning will need some time ...

Thanks
Right. The numbers above are for the 2600pgn book + my own modifications of lines that I think YAce is lacking to form what i call 2600b.pgn book..

I did it three times.
Sometimes it was faster than others..But It seems to vary with the entry parimeter X (the book_create X 20 ply) , used..
But no more than 10 minutes..

Sure thanks. try emailing it to s7716379d@ntu.edu.sg.
I assume future versions of Yace, won't have this bug as well?

The point is you still need to spend time recreating a book..And for large book it takes time..


Regarding the use of comments. Does it work currently with Yace020?
I created a file aaron.pgn
with 3 lines
1 e4 c5!
1 e4 c5!
1 e4 c5!
Then, followed the bookc.inp procedure you mentioned before, to combine with my gigantic 112 MB book.pgn file..
After the book was made, I found that Yace didn't play the sicilian..


Yes..
Thanks. For some reason i like matching programs with "learning".. since it gives me a illusion that I'm somehow making the engine stronger with each game i play..