The CGA (Chess Game Analyzer, available here: CGA on github) module produces QARC (Quantitative-Analytic Reports on Chess): entirely computer generated book-length LaTeX reports that turn a PGN collection into a reproducible, data-backed “game dossier.” The attached sample compiles Hans Berliner’s 16 games from the 5th World Correspondence Championship (ICCF 1965, WCC championships on wikipedia) into a single LaTeX book generated from Stockfish 17.1 analysis, pairing traditional annotation with interpretable engine-derived positional metrics.

Each game is presented as its own chapter with: (1) clean game metadata and player statistics (accuracy, average centipawn loss, and move-quality counts), (2) an annotated move-by-move score, (3) a critical positions section with FEN-based diagrams and “instead of / best continuation” recommendations, and (4) a quantitative layer of plots for evaluation, Space, Mobility, King Safety, and Threats, plus a composite Fireteam Index that attempts outcome prediction via sustained-post-opening dominance (raw and smoothed variants). The document also includes a collection-level prediction summary and a methodology appendix explaining how the underlying metrics are computed (in classical-eval terms for interpretability). Finally, for downstream research (e.g., regression/classification), the QARC output appends raw per-ply positional data for every game (SAN, centipawn eval, and all metric components for both sides) in a machine-friendly comment-block format.

For more info on CGA and QARC, see Medieval monks to modern metrics.