And now, for the criminally insane:

[Dayffd]

Ahh, FobbleSnoof. She's the best.

Maybe so, but I've always enjoyed the complications, ultimately unsound combinations and risks MightySnail would take, making for exciting games.

[lmader]

Ahh, FobbleSnoof. She's the best.

Maybe so, but I've always enjoyed the complications, ultimately unsound combinations and risks MightySnail would take, making for exciting games.

Boy howdy, when MightySnail takes the helm, the wood really starts flying! Oh my, good times.

[mariaclara]

thanks Dann.......

Well over 100 million FICS chess games, carefully filtered for dups, in SCID format, and compressed with bzip2:

[beachknight]

thanks Dann.......

Well over 100 million FICS chess games, carefully filtered for dups, in SCID format, and compressed with bzip2:

Hey Maria,

Welcome here

Best,

[Dann Corbit]

Btw, anyone knows any data mining projects with chess? I mean not endgame tablebase mining but something that could be done on such game-databases?

The usefulness might be less than you think.

All of the games are rated games, but...
The vast majority of the games will be things like:

White: MightySnail Elo 1475
Black: FobbleSnoof Elo 1613

Imagine mining data and finding that, say, for lower rated players Knight > Bishop or that two bishops do not value as much as for higher rated players. Wouldn't this be interesting?

I imagine that the data may have great value for chess teaching systems.
We can find out:
What sort of errors do players of level X make?
Here we have millions upon millions of rated games by humans. We can analyze for blunders and brilliancies and then find out what sort of players make what sort of mistakes or great moves.