🎾 Tennis Match Predictor 🎾

Calibrated Elo + ranking model, honestly evaluated (holdout: 64% accuracy, AUC 0.70, well-calibrated). Tennis is genuinely hard to predict — even bookmakers only reach ~70% — so treat these as informed estimates, not betting advice.

📝 Quirks & fine print

🧊 Retired or inactive players (yes, Roger Federer is in the list) keep the rating and ranking they had at their last match — frozen in time, so fantasy matchups come with extra salt.
📅 Ratings know nothing before 2020 and nothing after the last data refresh — recent hot streaks may be missing.
🎾 Only players with tour-level ATP matches since 2020 are rated (~635) — no Challengers, juniors, or WTA (yet?).
✍️ Names follow the dataset's spelling — no accents or hyphens ("Felix Auger Aliassime") — the autocomplete has the exact forms.
🎲 A 62% favorite still loses 38% of the time. That's tennis, and the whole point of a calibrated model.