Tennis track record
Every tennis prediction is logged before the match and graded afterwards. Nothing is deleted. NFL track record
Live picks
The live tennis record starts with the first matches after launch. Every pick is logged before the match and graded after it, wins and losses alike. A retirement counts for the player who went through; a walkover isn't graded.
Backtest, 2014 to 2026
Each year was predicted using only matches before it: ATP and WTA singles from ESPN (tour level, Grand Slams, WTA 125s and qualifying), walkovers and retirements left out.
ATP (men)
| Winners picked | Log loss | |
|---|---|---|
| AZMango, 2020 on (23,981 matches) | 65.3% | 0.6171 |
| Rating alone (Elo), same matches | 64.5% | 0.6270 |
| AZMango, main draws 2020 on (17,600) | 66.1% | 0.6080 |
| AZMango, 2014 on (39,762) | 66.4% | 0.6077 |
Total games, 22,625 matches in the last six seasons: our expected total was off by 5.51 games on average, against 5.56 for the tour's usual total on the surface. Straight sets or not: log loss 0.659 against 0.662 for the usual rate. Game margin off by 4.11 games on average.
WTA (women)
| Winners picked | Log loss | |
|---|---|---|
| AZMango, 2020 on (31,031 matches) | 66.3% | 0.6110 |
| Rating alone (Elo), same matches | 64.7% | 0.6231 |
| AZMango, main draws 2020 on (22,791) | 66.2% | 0.6109 |
| AZMango, 2014 on (46,134) | 66.0% | 0.6133 |
Total games, 29,805 matches in the last six seasons: our expected total was off by 4.86 games on average, against 4.91 for the tour's usual total on the surface. Straight sets or not: log loss 0.635 against 0.638 for the usual rate. Game margin off by 4.62 games on average.
The honest summary: our ratings pick about two winners in three, a little better than a plain Elo rating. We could not test them against the closing betting line: no free source of past tennis odds allows automated use, so we don't claim to match the market, let alone beat it. Published studies put the closing line at about 70% of ATP winners, a few points above us. Where we have a sportsbook line, the pages show the market's chance; the record above will show how we compare from now on.
Are the percentages honest?
| We said | Matches | Average call | Actually won | |
|---|---|---|---|---|
| ATP 50%-60% | 13,293 | 55% | 54% | |
| ATP 60%-70% | 11,437 | 65% | 64% | |
| ATP 70%-80% | 8,415 | 75% | 74% | |
| ATP 80%-90% | 4,862 | 84% | 82% | |
| ATP 90%-100% | 1,755 | 94% | 92% | |
| WTA 50%-60% | 17,010 | 55% | 55% | |
| WTA 60%-70% | 14,075 | 65% | 65% | |
| WTA 70%-80% | 9,640 | 75% | 75% | |
| WTA 80%-90% | 4,702 | 84% | 85% | |
| WTA 90%-100% | 707 | 92% | 92% |
Year by year
| Year | Matches | Winners picked | Log loss |
|---|---|---|---|
| ATP 2014 | 2,531 | 69.7% | 0.5722 |
| ATP 2015 | 2,596 | 70.0% | 0.5621 |
| ATP 2016 | 2,668 | 70.1% | 0.5804 |
| ATP 2017 | 2,630 | 66.8% | 0.6023 |
| ATP 2018 | 2,692 | 65.2% | 0.6253 |
| ATP 2019 | 2,664 | 66.1% | 0.6166 |
| ATP 2020 | 1,356 | 65.6% | 0.6136 |
| ATP 2021 | 2,699 | 65.5% | 0.6083 |
| ATP 2022 | 4,605 | 67.7% | 0.5992 |
| ATP 2023 | 3,941 | 63.5% | 0.6297 |
| ATP 2024 | 4,027 | 64.7% | 0.6207 |
| ATP 2025 | 3,676 | 64.4% | 0.6267 |
| ATP 2026 | 3,677 | 65.7% | 0.6201 |
| WTA 2014 | 2,378 | 66.1% | 0.6120 |
| WTA 2015 | 2,507 | 65.0% | 0.6152 |
| WTA 2016 | 2,571 | 66.9% | 0.6100 |
| WTA 2017 | 2,471 | 64.1% | 0.6303 |
| WTA 2018 | 2,575 | 65.2% | 0.6210 |
| WTA 2019 | 2,601 | 65.4% | 0.6204 |
| WTA 2020 | 1,225 | 64.0% | 0.6210 |
| WTA 2021 | 3,008 | 67.0% | 0.6033 |
| WTA 2022 | 5,177 | 67.4% | 0.6060 |
| WTA 2023 | 4,854 | 66.4% | 0.6105 |
| WTA 2024 | 5,162 | 66.0% | 0.6146 |
| WTA 2025 | 5,554 | 65.2% | 0.6173 |
| WTA 2026 | 6,051 | 66.5% | 0.6087 |
Data: ESPN's public tennis scoreboard (draws, scores, results); lines from The Odds API.