ATP and WTA · walk-forward

Tennis

Surface Elo calls the match winner. A serve and return simulation prices the set spread and the total games. There is no upcoming board: no free schedule feed is wired, so this page is the 2015–2025 backtest and a few settled examples.

Winner, all matches

63.8%

58,883 matches · log loss 0.632

Winner vs favorite

62.9%

Favorite 68.0% on 16,020 priced matches

Log loss vs the close

0.639

Market 0.589 · lower is better

The model does not beat the de-vigged closing favorite. Set spread and total games are computed for every match and are not graded: the free odds are match-winner prices only. tennis-data.co.uk, which publishes those winner prices and not a set or game line, was behind a Cloudflare challenge and was not used. The favorite comparison uses Valuebet closing prices for 2022–2025. The 2021 file has no close.

By season

SeasonMatchesModelPricedModelFavoriteLog lossMarket
20155,56365.6%—————
20165,82365.0%—————
20175,74064.4%—————
20185,80063.3%—————
20195,34163.5%—————
20202,72364.8%—————
20215,28263.6%—————
20225,46663.6%330663.1%67.5%0.6390.589
20235,75263.0%446563.2%68.3%0.6380.586
20245,91062.9%381962.9%68.3%0.6370.581
20255,48362.7%443062.6%67.6%0.6400.599

Tour

ATP

Model

64.5%

Favorite

68.1%

WTA

Model

63.1%

Favorite

67.8%

Surface

Hard

Model

64.8%

Favorite

67.4%

Clay

Model

62.8%

Favorite

69.0%

Grass

Model

60.9%

Favorite

68.1%

Settled examples

The last matches in the backtest. The probability is the surface Elo chance the eventual winner was ahead before the match. Expected games and the straight-sets chance come from the serve simulation. They are not a bet.

ATP · Hard · Next Gen FinalsModel had the loser

Alexander Blockx def. Justin Engel

1234Blockx3(7)444Engel4222

Winner prob

44.5%

Straight sets

16.0%

Expected games

42.2

Favorite

Won

ATP · Hard · Next Gen FinalsModel had the loser

Alexander Blockx def. Nishesh Basavareddy

123Blockx444Basavareddy3(2)3(8)1

Winner prob

34.6%

Straight sets

17.4%

Expected games

41.9

Favorite

Won

ATP · Hard · Next Gen FinalsModel had the loser

Alexander Blockx def. Dino Prizmic

1234Blockx4244Prizmic3(4)420

Winner prob

48.6%

Straight sets

14.2%

Expected games

41.7

Favorite

Won

ATP · Hard · Next Gen FinalsModel had the winner

Nicolai Budkov Kjaer def. Rafael Jodar

1234Kjaer4414Jodar1242

Winner prob

51.2%

Straight sets

11.3%

Expected games

41.4

Favorite

Lost

ATP · Hard · Next Gen FinalsModel had the winner

Rafael Jodar def. Martin Landaluce

123Jodar444Landaluce3(7)13(2)

Winner prob

51.8%

Straight sets

18.4%

Expected games

41.0

Favorite

Won

ATP · Hard · Next Gen FinalsModel had the winner

Nicolai Budkov Kjaer def. Martin Landaluce

1234Kjaer43(7)44Landaluce1423(4)

Winner prob

58.4%

Straight sets

19.1%

Expected games

41.2

Favorite

Lost

ATP · Hard · Next Gen FinalsModel had the loser

Rafael Jodar def. Learner Tien

12345Jodar14144Tien43(3)423(4)

Winner prob

17.0%

Straight sets

16.1%

Expected games

41.4

Favorite

Lost

ATP · Hard · Next Gen FinalsModel had the winner

Learner Tien def. Nicolai Budkov Kjaer

1234Tien3(2)444Kjaer4122

Winner prob

77.2%

Straight sets

10.6%

Expected games

41.6

Favorite

Won

ATP · Hard · Next Gen FinalsModel had the winner

Learner Tien def. Martin Landaluce

1234Tien1444Landaluce413(4)3(2)

Winner prob

84.6%

Straight sets

18.3%

Expected games

41.5

Favorite

Won

ATP · Hard · Next Gen FinalsModel had the loser

Alexander Blockx def. Nicolai Budkov Kjaer

123Blockx444Kjaer3(4)3(8)2

Winner prob

46.2%

Straight sets

14.1%

Expected games

41.9

Favorite

Won

ATP · Hard · Next Gen FinalsModel had the winner

Learner Tien def. Nishesh Basavareddy

123Tien444Basavareddy213(3)

Winner prob

76.5%

Straight sets

15.6%

Expected games

41.9

Favorite

Won

ATP · Hard · Next Gen FinalsModel had the winner

Learner Tien def. Alexander Blockx

123Tien444Blockx3(4)21

Winner prob

80.6%

Straight sets

9.7%

Expected games

41.8

Favorite

Won

Match data by Jeff Sackmann / Tennis Abstract, CC BY-NC-SA 4.0. Closing prices by Valuebet, CC BY 4.0. Eick Bet does not take bets.