Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is Mega Database in danger of becoming FatBase?

Author: Norm Pollock

Date: 11:39:35 01/13/05

Go up one level in this thread


On January 13, 2005 at 12:58:47, David Dahlem wrote:

>Hi Norm
>
>I'm currently downloading your pgn database from Peter Skinners site. Could you
>provide details on how you filtered this database?
>
>Thanks
>Dave

Games are from jan 1, 2000 to present. Sources are TWIC and ChessCollect

Filtered out (as best as possible) using Scid 3.61 and pgn-extract: players
under 2400 elo, games with [FEN (includes FRC), blindfold, blitz, rapid,
lightning, simultaneous, email, twins and duplicates, computer engines, 20 moves
(40 plies) or less, games on the Internet (ICC, FICS, playchess, IEC), unusual
ratings or unusual names.

And of course you have the option of filtering out even more.

Stripped out excess tags.

Some suggestions for book builders:
Use a high number of occurrences if you are building a book. I recommend 12. Do
not use book learning if multiple engines share the book. The two book approach-
a white book and a black book is better than the 1 book approach. That is
because you can build the white book just from games that white wins/draws, and
build the black book just from games that black wins/draws.





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.