Author: José Carlos
Date: 01:33:20 01/16/01
Go up one level in this thread
On January 15, 2001 at 14:34:26, Dann Corbit wrote:
>On January 15, 2001 at 07:47:13, José Carlos wrote:
>
>>On January 13, 2001 at 16:39:31, Dann Corbit wrote:
>>
>>>On January 13, 2001 at 14:52:08, Leen Ammeraal wrote:
>>>
>>>>On January 13, 2001 at 14:35:49, Pete Galati wrote:
>>>>
>>>>>On January 13, 2001 at 13:46:15, Leen Ammeraal wrote:
>>>>>
>>>>>>I just uploaded a (rather restricted) WinBoard version,
>>>>>>wbqueen, of my chess program queen. Any reactions
>>>>>>are welcome. Being unfamiliar with ELO rating I would
>>>>>>appreciate any opinions (or better) about the rating
>>>>>>of this program, as well as suggestions how to improve it.
>>>>>>I programmed it in VC++6.0 and I used
>>>>>>
>>>>>>- rotated bitboards
>>>>>>- PVS/Negascout
>>>>>>- hashing
>>>>>>- nullmove pruning
>>>>>>- openings from the PGN files of about 350 FIDE games
>>>>>>
>>>>>>Leen Ammeraal
>>>>>>http://home.wxs.nl/~ammeraal/
>>>>>
>>>>>It always surprizes me when someone builds the opening book inot their exe. It
>>>>>seems to be working just fine, I didn't take the time to add it to the WBini
>>>>>file yet, but it's working without any problems so far.
>>>>>
>>>>>When you say "rather restricted", do you mean it doesn't use many Winboard
>>>>>commands and options yet?
>>>>>
>>>>>Pete
>>>>
>>>>Yes, that is what I mean. As for the opening book, it may interest
>>>>you to see a code fragment related to it:
>>>>
>>>>char *pOpenings[] =
>>>>{"e4 e5 Nf3 Nc6 Bc4 Bc5 c3 Nf6 d3 a6 Nbd2 O-O Bb3 d6 h3 Be6 Nf1 h6 g4 Nh7 Rg1 d5
>>>>Qe2 Be7 ",
>>>>"d4 Nf6 Nf3 g6 c4 Bg7 g3 d6 Bg2 O-O O-O Nc6 Nc3 a6 d5 Na5 Nd2 c5 Rb1 Rb8 b3 b5
>>>>Bb2 Bf5 ",
>>>>"Nf3 Nf6 d4 e6 c3 c5 Bg5 cxd4 cxd4 Qb6 Qc2 Nc6 Bxf6 gxf6 e3 d5 Nc3 Bd7 Be2 Rc8
>>>>O-O f5 Rfc1 Nb8 ",
>>>>"g3 d5 Bg2 g6 c4 c6 cxd5 cxd5 Qb3 Nf6 Nc3 e6 Nf3 Nc6 d4 Bg7 O-O O-O Bf4 Nh5 Rfd1
>>>>Nxf4 gxf4 b6 ",
>>>>
>>>>These are the first four of the 350 games I used. I downloaded
>>>>some file with these games from the Internet, and made them
>>>>somewhat shorter by deleting the move numbers and tags (using a simple
>>>>conversion program, which generated the above code). All positions
>>>>that occur with these moves, along with the next move, are stored
>>>>in my hashtable, using a modular counter that start at a random
>>>>position in this array pOpenings (to avoid getting the same
>>>>game each time), and that is the opening book! If you think
>>>>there is a better set of games for this purpose, please let me know.
>>>
>>>The Winboard Forum is a good source of information for this sort of thing also:
>>>http://f11.parsimony.net/forum16635/
>>>
>>>Mogens Larsen seems to have a knack for building good books. The very best
>>>books are made by the professionals. I suspect that Vincent has a really good
>>>book because he puts a lot of effort into it. Jeroen N. is a professional book
>>>builder (If I recall correctly).
>>>
>>>There is a PGN file on my ftp site that contains exclusively players with an ELO
>>>rating of 2600 or more on both sides of the board. These games are obviously of
>>>very high quality.
>>>ftp://cap.connx.com/pub/2600.ZIP
>>
>> Hi Dann, I tried your pgn some time ago and I find it really interesting. The
>>games it contains are not perfect, and there're still several blunders by the
>>involved GM's there, but it's a good starting point for building an own book.
>> I've deleted some games from it, included '!' and '?' marks here and then, and
>>pasted some analisys for building a new book for my program. Right now, I'm
>>testing it...
>> But, just an advice, don't build a book from that pgn without some checking,
>>or the program will lose some stupid games...
>>
>> Anyway, thanks for the file.
>>
>> José C.
>
>If you can provide the updated information, that may prove very helpful. Are
>you using NCO to correct it or ???
I'll take a look at it when back home (at work, right now), but I don't think
my work here is valuable for others.
First of all, I removed blunders. This could be useful, but I didn't keep that
"version" of the file. BTW, I used "game-testing" with Averno against strong
programs to find the blunders, and got no idea what NCO is.
Secondly, I flagged as '?' the openings moves I didn't want Averno to play,
and deleted some games that were not interesting for me. For example, if I flag
1. f4? then I delete all games starting with 1. f4 that white wins.
In the third place, I compiled the book, played games and inserted analisys on
the openings played in the games, to improve Averno's play in the opening. Then
played games again and so on...
So, I'm afraid this is not useful for other people. But anyway, I can try to
send it to anyone, if interested... no big secrets in Averno's book :)
But are you ready for almost 30Mb? ;)
José C.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.