FAQ
Questions, answered
Is ChessLab really free?
Yes. It is free, with no ads and no account. It is a one-person project, and the source code is open.
Does it need an internet connection?
No. Everything works offline, in airplane mode: the engines, the analysis, the courses and the puzzles are all inside the app. The only thing that can use the network is the optional iCloud sync, which is off by default.
What happens to my games and progress?
They stay on your device, in the app's private storage. You can optionally sync them between your own iPhone and iPad through iCloud (off by default); they then live in your private iCloud space, which we cannot access.
Can I use it on several devices?
Yes. With iCloud sync switched on, your games, progress and repertoires follow you between your own iPhone and iPad. It is optional and off by default. Nothing is stored on any server of ours: we have none.
How do the human-like opponents work?
Nine characters are played by Maia-3, a neural network from the University of Toronto trained on millions of human games. It does not look for the best move: it plays the move a human of that level would play, mistakes included. Each character adds a style — attack, trades, gambits — without ever buying itself a blunder the model found unlikely.
What level are the opponents?
They use the human scale of Maia, close to the Lichess scale. It is not comparable with a Stockfish level: a Stockfish throttled to 1100 beats a Maia set at 2200, because it calculates where Maia recognises. Two scales, two different things. You can also play Stockfish itself, from about 900 to about 3190 Elo.
Which variants are included?
Chess960 and eleven others against the computer: King of the Hill, Three-Check, Horde, Atomic, Crazyhouse, Duck Chess, Barricades and more, each with its own analysis. There is also a two-player mode on one device.
Can I import my own games, repertoires and puzzles?
Yes. Load a game or a collection of games from a PGN file, paste a PGN or a FEN, build a position in the editor, or scan one from a photo. You can import your own opening repertoires as PGN, and your own puzzles as PGN files, which become collections you can drill, including with the Woodpecker method.
How does the scanner work?
Photograph a real board, or open a screenshot, and the position lands on the board. The recognition runs on your device and the image is sent nowhere. Whose turn it is, castling rights and en passant cannot be read from a picture, so the app does not guess: you set them.
How does the spaced repetition work?
Openings use FSRS-5, puzzles use SM-2. The unit is the position, not the line: a position you learned in one opening counts in every opening that reaches it. A daily session tells you what is due.
Which languages are supported?
French, English, Spanish, German and Italian, fully, including the opening and endgame courses and the coaching comments. The piece letters follow the language (K, Q, R… or R, D, T…), or you can choose them in the settings.
Why is it open source?
ChessLab embeds Stockfish, which is released under the GPLv3 licence. That licence requires the complete source code to be available, so it is, and so are the notices for the other components.
I found a bug, or I have an idea.
Write to thmaed@icloud.com or open an issue on GitHub. It is a one-person project with no guaranteed response time, but everything is read.