Calculation Training is made by Say Chess, and can be reached at saychess1@gmail.com. It currently holds 3,142 verified positions.
Game material is mined from The Week in Chess by Mark Crowther — four decades of weekly chess journalism; consider donating. Verification and grading run on Stockfish (GPLv3). Human findability is modelled with Maia-2 (MIT), the skill-conditioned human-move model from the Computational Social Science Lab at the University of Toronto — Maia-2: A Unified Model for Human-AI Alignment in Chess, NeurIPS 2024. Mining also reads one search-free evaluation per position from Leela Chess Zero (GPLv3). Board legality and rendering use python-chess (GPLv3). The two neural models run offline, while the bank is being built, and never on the host serving this page; Stockfish is the one engine that also runs live, probing your line when you commit it.
| set | author | licence |
|---|---|---|
| classic / ivory (cburnett) | Colin M. L. Burnett | GPLv2+ |
| merida | Armando Hernández Marroquín | GPLv2+ |
| leipzig | Armando Hernández Marroquín | freeware |
| chessnut | Alexis Luengas | Apache 2.0 |
| celtic | Maurizio Monge | MIT |
Piece artwork is distributed from lichess-org/lila under the licences above. The share buttons’ brand glyphs are from Simple Icons (CC0).