Win probability
Sunday, 12 July 2026
5 matches selected · 5 tournaments · model last ran 16:50 UTC
Umpiry rates 5 matches on Sunday, 12 July 2026. The closest is Mayar Sherif v Jeline Vandromme at 59 % — the model’s most uncertain call of the day. The most one-sided is Jannik Sinner v Alexander Zverev at 88 %. Probabilities come from a per-tour model trained on Elo, surface Elo, recent serve and return form and layoff.
| Time | Tournament | Round | Match | Probability | Verdict | H2H |
|---|---|---|---|---|---|---|
| 13:35 | Contrexeville CHALLENGER · clay | F |  Sherif M. Vandromme J. | 59 %41 % | Slight edge | 1–0 |
| 17:10 | Newport CHALLENGER · grass | F |  Volynets K. Maria T. | 61 %39 % | Slight edge | 3–2 |
| 12:20 | Braunschweig CHALLENGER · clay | F |  Choinski J. Gaston H. | 64 %36 % | Slight edge | 1–0 |
| 11:10 | Nottingham 4 CHALLENGER · grass | F | Searle H.Chidekh C. | 65 %35 % | Slight edge | 1–1 |
| 15:10 | Wimbledon ATP · grass | F |  Sinner J.Zverev A. | 88 %12 % | Strong favourite | 10–4 |
| Time | Tournament | Round | Match | Probability | Verdict | H2H |
|---|---|---|---|---|---|---|
| 15:10 | Wimbledon ATP · grass | F | Sinner J.Zverev A. | 88 %12 % | Strong favourite | 10–4 |
How to read this
Each row gives the model’s probability that the favourite wins, and the verdict says the same thing in words. Rows are ordered by how close the model thinks the match is, so the top of the table is where it is least sure. A 70 % favourite losing is not a broken model — it is the 30 %. Probabilities are not odds and take no account of any market.
How often the model has been right
| Tour | Correct |
|---|---|
| ATP / WTA | 66 % |
| Challenger (men) | 64 % |
| Challenger (women) | 68 % |
| ITF (men) | 66 % |
| ITF (women) | 64 % |
Measured on the predictions the model actually made, not on a backtest — see Methodology.
Yesterday
Saturday, 11 July 2026 · 5 matches
Favourites won
2 / 5
Model expected
2.9
The model expected 2.9 favourites to win; 2 did. See yesterday’s page.
Biggest shock, last 7 days
Questions
- How is the probability calculated?
- A separate model for each tour, trained on Elo, surface-specific Elo, recent serve and return form and time since the player’s last match. It runs several times a day; the time of the last run is at the top of this page.
- How often is the model right?
- We publish it. The table above gives the hit rate for each tour family, measured on the predictions the model actually made — see Methodology for how it is computed.
- Why are the closest matches at the top?
- Rows are ordered by distance from 50 %, so the table opens where the model is least sure. “Most one-sided” reverses the order and groups the rows by tournament.
- Is this a betting tip?
- No. The page shows a model’s probability and how often that model has been right. There are no odds, no comparison with a market and no suggested stake.