For tennis quants, model builders, and serious bettors who asked how we calibrate the Lob model: we have packaged our complete historical research database and out-of-sample probability benchmark into a standalone download ($39).
What is inside the v1.0.0 bundle (SQLite, Apache Parquet, CSV, 5-Sheet Excel Workbook, and Python 3 Backtest Scripts):
• 33,136 Historical ATP Matches (2015–2026): 12 full seasons of tour-level & Grand Slam match results with 50 standardized box-score columns (surface, indoor/outdoor, player rankings, serve %, aces, double faults, 1st/2nd serve points won, break points saved/faced, duration).
• 1,214 ATP & WTA Player Surface Ratings: Surface-specific Elo (Overall, Hard, Clay, Grass) merged with Glicko-2 Rating, Rating Deviation (RD), and Volatility.
• Out-of-Sample Model vs. No-Vig Market Ledger (253 matches): Every pre-match Lob win probability alongside the de-vigged bookmaker consensus probability, edge in bps, best decimal price, Brier Score, Log-Loss, Quarter-Kelly sizing, and $10 flat PnL trajectory (+$127.20 / +14.13% ROI on ATP core picks <= (8.00)).
Dataset overview, schema & instant download ($39): https://lobdata.app/dataset/?utm_source=covers&utm_medium=marketplace&utm_campaign=dataset39
18+ · Quantitative research & educational data product. Bet responsibly.
