A model that generates its own win probabilities for ATP/WTA matches from match history alone — Elo, form, head-to-head, surface and venue splits — then checks itself against real bookmaker closing lines. No market data goes into the predictions; the market is only used here to grade us.
All numbers below are on matches from 2024 onward that were never used in training — a true out-of-time holdout, not a lookback.
| Accuracy | Log loss | Brier | n | |
|---|---|---|---|---|
| Elo only | 65.6% | 0.6101 | 0.2118 | 14,150 |
| Our model | 66.9% | 0.5964 | 0.2061 | 14,150 |
| Real market (Pinnacle, de-vigged) | 68.3% | 0.5885 | 0.2023 | 11,052 |
Matched 11,052 / 14,150 main-tour 2024+ matches to closing odds (78% coverage). We pick the same favorite as the market 90% of the time.
When a predictor says "70% to win," does that side actually win about 70% of the time? The closer a line sits to the diagonal, the more trustworthy its numbers are.
Live predictions for upcoming matches are coming soon — this needs a fixtures feed wired in (in progress). For now, everything above is measured against past matches the model never trained on.
Feature ideas, a stat you'd want tracked, a bug in the numbers — all welcome.