A small offer.

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 26 July 2000 11:15:28:
If anyone is contemplating making similar opening books for different programs, I have a small offer to make. I've collected 42500 games (includes 2600.pgn) without duplicates and uploaded them to
http://www.geocities.com/mogens72/Games/
The format is .cbv, so a chessbase product capable of accepting that number of games is necessary (no CBLight). There's also a file with flags for use with the upcoming Yace 0.22 version.
Best wishes...
Mogens

[Winfred Craft]

Geschrieben von: / Posted by: Winfred Craft at 26 July 2000 22:47:55:
Als Antwort auf: / As an answer to: A small offer. geschrieben von: / posted by: Mogens Larsen at 26 July 2000 11:15:28:

If anyone is contemplating making similar opening books for different programs, I have a small offer to make. I've collected 42500 games (includes 2600.pgn) without duplicates and uploaded them to
http://www.geocities.com/mogens72/Games/
The format is .cbv, so a chessbase product capable of accepting that number of games is necessary (no CBLight). There's also a file with flags for use with the upcoming Yace 0.22 version.
Best wishes...
Mogens

That could be interesting for my opening book but I have to be sure your book
contains only full games between computer-computer,
so no human or human-computer games.
Drop me a few lines, Mogens
wincraft5@hotmail.com
Bye,
Winfred

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 27 July 2000 00:51:25:
Als Antwort auf: / As an answer to: offer geschrieben von: / posted by: Winfred Craft at 26 July 2000 22:47:55:

That could be interesting for my opening book but I have to be sure your book
contains only full games between computer-computer,
so no human or human-computer games.

There are no computer-computer games in the collection I've published. If you want that kind of games, there are a lot of sources. From the top of my head I would suggest Frank Quisinsky's collection or maybe games from Chessfun's chess pages at http://www.geocities.com/chessfun_1999/. I have a few games available for download as well on my WinBoard page. There's the SSDF data and the Cadaque tournament (I can't remember this link), so the sources are plentyful.
Best wishes...
Mogens

SSDF

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 27 July 2000 01:02:22:
Als Antwort auf: / As an answer to: Re: offer geschrieben von: / posted by: Mogens Larsen at 27 July 2000 00:51:25:

That could be interesting for my opening book but I have to be sure your book
contains only full games between computer-computer,
so no human or human-computer games.

There are no computer-computer games in the collection I've published. If you want that kind of games, there are a lot of sources. From the top of my head I would suggest Frank Quisinsky's collection or maybe games from Chessfun's chess pages at http://www.geocities.com/chessfun_1999/. I have a few games available for download as well on my WinBoard page. There's the SSDF data and the Cadaque tournament (I can't remember this link), so the sources are plentyful.

There is something recursive about his request.
The opening books generated by a computer against a computer where neither has an opening book are pathetic, for the most part. Hence, all computer opening books must (first) be initially generated by an influx of human-human games.
Now that the computer has an opening book generated from such games, the openings it spews out are -- you guessed it -- exactly those that are played by humans.
If you remove the opening books, some of the best programs will come up with decent (not generally great) moves at 40/2, but they will clearly be inferior to that generated by GM's.
Perhaps he wants the openings checked for blunders. Though incredibly rare if both players are GM's and even less likely if we force that the same sequence is repeated multiple times, it is conceivable that a blunder would sneak through. The solution to that problem is to preanalyze the chosens positions for blunders and tag them and/or to use learning to play/avoid winning/losing lines.


my ftp site

[The co-whiner is no other]

Geschrieben von: / Posted by: The co-whiner is no other than Mogens Larsen at 27 July 2000 01:42:50:
Als Antwort auf: / As an answer to: Re: offer geschrieben von: / posted by: Dann Corbit at 27 July 2000 01:02:22:

The opening books generated by a computer against a computer where neither has an opening book are pathetic, for the most part. Hence, all computer opening books must (first) be initially generated by an influx of human-human games.
If you remove the opening books, some of the best programs will come up with decent (not generally great) moves at 40/2, but they will clearly be inferior to that generated by GM's.
Perhaps he wants the openings checked for blunders. Though incredibly rare if both players are GM's and even less likely if we force that the same sequence is repeated multiple times, it is conceivable that a blunder would sneak through. The solution to that problem is to preanalyze the chosens positions for blunders and tag them and/or to use learning to play/avoid winning/losing lines.

Yes, and most are told what moves to select, because human experience judge them as being objectively better at this point in time. But there will be unique computer moves at some depth, even though they are few compared to the total count.
Many programs produce silly opening moves when left to their own devices. Funny enough, Yace actually produces the Petroff when confronted with the moves e4 and Nf3. I noticed that when doing a little larsen_mode test. It'll even play e5 on it's own.
Very few blunders are repeated several times within the first 30ply, which is the usual max book depth in computer games, and therefore rarely selected. If the blunders are in the book, they will be repeated by computers.
Best wishes...
Mogens

[W.Craft]

Geschrieben von: / Posted by: W.Craft at 27 July 2000 15:17:22:
Als Antwort auf: / As an answer to: Re: offer geschrieben von: / posted by: Mogens Larsen at 27 July 2000 00:51:25:

That could be interesting for my opening book but I have to be sure your book
contains only full games between computer-computer,
so no human or human-computer games.

There are no computer-computer games in the collection I've published. If you want that kind of games, there are a lot of sources. From the top of my head I would suggest Frank Quisinsky's collection or maybe games from Chessfun's chess pages at http://www.geocities.com/chessfun_1999/. I have a few games available for download as well on my WinBoard page. There's the SSDF data and the Cadaque tournament (I can't remember this link), so the sources are plentyful.
Best wishes...
Mogens

Thanks for your info, Mogens
...of course, most moves from opening books were once played by GM...
...but that topic was not at issue.
Winfred

[Aaron]

Geschrieben von: / Posted by: Aaron at 27 July 2000 18:22:24:
Als Antwort auf: / As an answer to: Re: offer geschrieben von: / posted by: The co-whiner is no other than Mogens Larsen at 27 July 2000 01:42:50:

Many programs produce silly opening moves when left to their own devices. >Funny enough, Yace actually produces the Petroff when confronted with the moves e4 and Nf3. I noticed that when doing a little larsen_mode test. It'll even play e5 on it's own.

Not that funny, in my experience..Most programs like petroff when book is off..

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 27 July 2000 20:35:52:
Als Antwort auf: / As an answer to: Re: offer geschrieben von: / posted by: The co-whiner is no other than Mogens Larsen at 27 July 2000 01:42:50:
[snip]

Many programs produce silly opening moves when left to their own devices. Funny enough, Yace actually produces the Petroff when confronted with the moves e4 and Nf3. I noticed that when doing a little larsen_mode test. It'll even play e5 on it's own.

Some version of crafty (I can't remember which one) would always play the Dragon if started without a book against itself.


my ftp site

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 27 July 2000 20:50:35:
Als Antwort auf: / As an answer to: Re: offer geschrieben von: / posted by: Dann Corbit at 27 July 2000 20:35:52:

[snip]

Many programs produce silly opening moves when left to their own devices. Funny enough, Yace actually produces the Petroff when confronted with the moves e4 and Nf3. I noticed that when doing a little larsen_mode test. It'll even play e5 on it's own.

Some version of crafty (I can't remember which one) would always play the Dragon if started without a book against itself.

For amusement I tried the usual d4, c4 and Nc3 development to see how Yace would respond. It choose the Nimzo indian.
[Event "Computer chess game"]
[Site "MCL"]
[Date "2000.07.27"]
[Round "-"]
[White "Mogens"]
[Black "Yace_022"]
[Result "*"]
[TimeControl "40/300"]
1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. a3 Bxc3+ 5. bxc3 O-O 6. e3 d6 7. Bd3 e5 8.
Ne2 e4 9. Bb1 Bg4 (stoppped)

Best wishes...
Mogens