Category: PGN Database

My latest project is to take the Lichess Open Directory (the full one) as well as the releases of the Chesscom Elite DB and filter them on the same criteria using the pgn-extract command line interface:

-t"tags.txt" -R"roster.txt" --xroster --minmoves 10 --maxmoves 250 --fixresulttags --fixtagstrings --nosetuptags -C -V -N -D -e --plycount -w9999

The “tags.txt” file contains:

WhiteElo >= "2300"
BlackElo >= "2300"
TimeControl >= "180"

and the “roster.txt” file contains:

Event
Site
Date
Round
White
Black
Result
WhiteElo
BlackElo
ECO
Opening
Variation
EventCountry
EventDate
SetUp
FEN
Variant
TimeControl
Beauty
PlyCount

Here are my finished online strong games:

https://chessnerd.net/pgn/lichess/lichess-strong

https://chessnerd.net/pgn/chesscom/cc-strong/

Also, here is a link to the Python script for extracting a PGN from a ZST archive, starting at 2300 elo. (That can be adjusted, of course.)

https://www.mediafire.com/file_premium/viyqko7e2rc33wt/zst-extractor.py/file

I don’t really have room to mirror the Chesscom Elite files, though I do have a complete set, which runs from January 2021 through to May 2024. The whole thing is about 130 GB. I did filter each file, however, and was able to put up a folder called Chesscom Elite Filtered over at MediaFire:

https://www.mediafire.com/folder/w0fn58vxmdsbn/Chesscom_Elite_Filtered

The set comprises the same 2021-01 to 2024-05, but has been filtered with pgn-extract to take out anything below a 2300 rating, or below 10 moves. To remove doubles and non-standard and bullet games, as well as strip comments and variations. Finally incorrect tags have been fixed and openings and plycount have been added. Uncompressed the whole thing is about 11 GB, though here it is compressed, and comprises more like 3 GB.

Games downloaded by chess forum user Splassky, using the Chesscom API.

CvC and CvH. Includes 1,492 different high-rated bots, as downloaded 2023-11-29. Filtered to include only beautiful strong games. Unannotated. Doubles removed. No variants. Beauty scores added.

439,985 games | 138 MB ZIP | 448 MB PGN
https://www.mediafire.com/file/n9cz54mnzclm0bg/lichess-bot-games.zip/file

Strong games are also referred to by ChessBase as good games, and top games. The criteria seem to be an average of 2300+ and 10+ moves.

Beauty is determined according to interesting play, such as sacrifices. Between 0-3 medals are awarded according to the score. Beautiful games, by my own definition, have anywhere from 1-3 medals.

These are match games organized by month, and training games organized by test run and by day.

Tags added for player, rating, event, site, event date, date, ECO, opening, beauty and PlyCount. All 960 games removed.

Lc0MG 2022-12 — 2,560,642 games | 2.41 GB PGN | 445 MB 7z
Lc0MG 2023-01 — 1,907,622 games | 1.80 GB PGN | 346 MB 7z

T60 2022-01-01 — 513,567 games | 490 MB PGN | 99 MB 7z
T60 2022-01-02 — 609,065 games | 581 MB PGN | 117 MB 7z
T60 2022-01-03 — 610,590 games | 583 MB PGN | 117 MB 7z
T60 2022-01-04 — 559,167 games | 499 MB PGN | 102 MB 7z
T60 2022-01-05 — 536,930 games | 482 MB PGN | 100 MB 7z
T60 2022-01-06 — 508,590 games | 455 MB PGN | 95 MB 7z
T60 2022-01-07 — 509,587 games | 456 MB PGN | 95 MB 7z