How opening books, mirrored openings and switched-side testing affect computer chess ratings and why pair-ratio discipline matters for public interpretation.