Credits

Kifubara would not exist without the collections and projects below. A heartfelt thank-you to the people who built and maintain them, and to the Go communities they came from.

GoGoD

Most of the historical games on Kifubara come from GoGoD, the Games of Go on Download database. T Mark Hall and John Fairbairn spent decades building it, buying rare collections and transcribing games by hand long before old records were easy to find online, and John still corrects and extends it as new discoveries surface in the Far East. When you replay a castle game here, or a title match from the 1950s, you are reading a record they saved. We share those games with their permission and with a daily download allowance at their request.

If you love the old games, the complete database is well worth owning, and John's books on Go history are the deepest writing on these games you'll find in English.

KataGo

Every win-rate graph, every flagged mistake, every "Play vs AI" bot move you see on Kifubara comes from KataGo, the open-source Go engine by David J Wu. We use both its strong network and its human-style network (humansl) — the latter is what lets us offer bots that play like real humans at a given rank rather than crushing every opponent at maximum strength. Kifubara is not affiliated with the KataGo project.

Shudan

The board you see on every game page — the wood, the stones, the coordinate labels — is rendered with Shudan, a Go board component by Yichuan Shen and the SabakiHQ contributors.

Board styles

Under the ⚙ menu on any game page you can pick what your board and stones look like. Credit goes to the creators below. We're grateful to them for publishing their work, for letting us all enjoy it, and for bringing the aesthetics of Go to computer screens.

• Goban Decals — illustrated boards, where the grid is painted into the artwork rather than drawn over it. By Borys Lykah, under CC BY-NC 4.0 and used here with his permission.
• Go Assets — the slate-and-clamshell stones you'll recognise from half the Go software on the internet, and the wood and photographed boards to go under them. By Andreas Tarnowsky, under CC BY-SA 4.0; the versions we serve are resized and repacked, and carry the same licence.
• Rendered boards — a whole shelf of boards and stones raytraced in Blender, in materials real gobans are and aren't made of. By billhails, SabakiThemes, MIT.
• Sabaki themes — glass, broadcast, vector and one that turns the board into a printed diagram. By upsided, Upsided-Sabaki-Themes, MIT.
• Subtle — ink-wash boards you play on top of a painting. By RobertChrist, Sabaki-Subtle-Theme, MIT.
• Subdued — for when you want the board to get out of the way. By fohristiwhirl, sabaki_subdued_theme_40, MIT.
• KaTrain — boards and stones from Sander Land's KaTrain, MIT.
• geovens — clean, flat boards, from Sabaki-Theme, shared for anyone to use in a Go app with credit.

Board sounds

The same menu lets you choose what sounds to use for stones. We have several available sets to choose from:

• Yunzi — a real Yunzi stone played on two real boards, a tabletop one and a thin shin-kaya, recorded and released into the public domain by billhails under CC0 on Freesound, with the note that it might come in useful to anyone building a software goban. He recorded no capture, so that set borrows Sabaki's below.
• Sabaki — stones and captures from Yichuan Shen's Sabaki, MIT.
• KaTrain — stones and a capture from Sander Land's KaTrain, MIT.
• Variant Go — a crisper stone and capture, by Jae from Variant Go Server, given to us for use here.

Preact

The interactive board surface — move navigation, variation exploration, the "Play vs AI" UI — is built with Preact, a small React-compatible framework by Jason Miller and the Preact contributors.

And the communities behind them

A separate thank-you to the KataGo distributed-training contributors whose donated GPU time produced the networks Kifubara runs on, and to the OGS and KGS communities — the human games played there are the source material the human-imitation network learned from. Without you, the bots wouldn't know how 8-kyu players actually play.