All Go games from online-go.com, distributed with permission of OGS. Contents: - 碁 (Go), 圍棋/围棋 (Weiqi), 바둑 (Baduk) games - 27,086,420 games. - Games are included from 2005-11-05 (start of OGS) to 2021-08-29 - Games marked private are omitted. - Games are provided as one SGF file each. - Files are always named {game_id}-{black}-{white}.sgf - sgfs-by-date.tar.gz: Files are organized by date (one folder per day) - sgfs-by-username.tar.gz: Files are organized by username (one folder per username). The 6 largest bot folders were deleted because Windows does not deal well with a million files per folder. human-bot games are still listed in the human player's folder. - sgfs-uploaded.tar.gz: Files are organized by date (one folder per day). Contains SGF files uploaded to OGS. These are typically from tournament or professional play. The date of organization is either the date of tournament play (where available) or the date of upload. - sgfs-by-date.tar.gz and sgfs-by-username.tar.gz contain the same non-tournament games. Games appear once each in sgfs-by-date, and twice each in sgfs-by-username (once for black and once for white). Neither includes the uploaded tournament games. - Games are also provided as one line of JSON each in ogs_games_2013_to_2021-08.json.gz - This is the raw API results for each game ID - Source code is included. If you want to submit a patch to better generated SGFs from the JSON (for example to fix the ranks), add a pull request to `https://github.com/lightvector/ogstosgf`. - Crawl period was 2021-08-29 to 2021-10-15 with permission of OGS. - Known issues - The ranks of the players are likely to be wrong in SGF files. - Chat is not included in this collection, but it's publicly available on online-go.com if you want it for a particular game. - sgfs-uploaded.tar.gz presents SGF files as originally uploaded. They may be missing information, incorrectly dated, or invalid files. This is expected, that's why they're in a separate collection. Can I use this? - Short answer: Yes - Longer answer: - Forum thread: https://forums.online-go.com/t/can-we-get-an-sgf-database-dump/38837 - I (Zachary Vance) release anything I have rights to into the public domain. lightvector doesn't have rights to anything here except the ogstosgf.py script, which they say use for whatever - anoek (as an OGS representative) has said it's fine to use this for research. I assume generally OGS should be pretty chill but feel free to ask them (NOT me) if you want to do something non-researchy - I haven't solicited permission from the millions of individual players who make up this collection, and I'm not going to. - I'm not a lawyer, it's unclear what the legal status of a collection of public games would even be, go bother a lawyer if you have legal questions. This collection is thanks to - OGS (especially anoek): for running a great online server for many years, making games available, giving permission, and assistance downloading games - Zachary "za3k" Vance: downloaded and put together collection, improved SGF conversion script - lightvector (author of KataGo): json to SGF conversion script