Our forecast record
81 forecasts scored since 2 November 2026, log loss 0.6689. The favourite won 61.7% of the time.
The record so far
| Forecasts written | 101 |
| Scored | 81 |
| Waiting for a result | 15 |
| Not counted | 5 |
| Log loss | 0.6689 |
| Brier score | 0.2367 |
| Main-draw matches played since the first forecast | 81 |
Log loss measures how much probability we gave to what happened: lower is better, and always saying 50% scores 0.6931.
Calibration
| Chance we gave the favourite | Matches | Average forecast | Favourite won |
|---|---|---|---|
| 50–55% | 16 | 52.3% | 62.5% |
| 55–60% | 18 | 56.6% | 61.1% |
| 60–65% | 6 | 61.7% | 16.7% |
| 65–70% | 11 | 66.6% | 45.5% |
| 70–75% | 11 | 72.2% | 81.8% |
| 75–80% | 10 | 76.9% | 90% |
| 80–85% | 6 | 82.5% | 50% |
| 85–90% | 2 | 86.3% | 50% |
| 90–95% | 1 | 92.6% | 100% |
Well-calibrated forecasts keep the last two columns close. A row with few matches moves a lot by chance.
How the record works
Before each main-draw ATP and WTA match we write our chance for each player. The database stamps the time and refuses any later change.
A forecast counts if the same two players finish the match, retirements included. A walkover, a cancelled match or a replaced opponent does not count.
The forecast comes from our own model: Elo ratings, serve and return skill estimated from match statistics, and age. How we built and tested it is in our methodology.
We do not claim these forecasts are better than anyone else's. This record is here so you can judge them. They are not betting advice.