Rethinking Women's French Open Seedings

It only took a few days of main draw play for the 2017 French Open to give fodder to debates on the unpredictability of this year’s women’s tour. After two rounds, just 9 of the 32 seeds survived to the round of 32, 23 fewer than expected. Those numbers caused tennis.com to conclude that the women’s draw was “decimated”.

Before the start of the French Open, I put forward the idea that a contributor to the apparent chaos on the women’s tour was the poor performance of the official rankings. Because of the peculiarities of the rankings, which use a point-based rather than prediction-based system, seedings of tournament draws in 2017 have repeatedly been flawed and caused an exaggerated sense that the “best players” aren’t living up to expectation.

In response to this idea, Jeff Sackmann of the Tennis Abstract pointed out that, while the WTA rankings have historically made far worse picks than alternative systems like Elo (especially surface-specific Elo), all methods would have struggled this year. Predictions from rankings and more probabilistic systems have all trended down this year, indicating that 2017 is inherently less predictable.

Thus, although rankings have added to the sense of a topsy-turvy women’s tour this year, they aren’t the only factor. Indeed, Sackmann suggests that what we are witnessing may be a transition period for the WTA, as the next generation fight for their place in a tour where the futures of previously dominant players, like Serena Williams and Victoria Azarenka, are uncertain.

If we are in a transitioning phase of the WTA tour, then how we treat historical performance could be especially crucial for predictions. If some players are on a steep upward trajectory while others are in the midst of a sharp decline, performance 12 months ago or even 6 months ago might have little value in setting our expectations today.

This made me curious to look at how much win expectations for the current season would vary with the time period of match history considered. So, I derived a surface-weighted Elo and their implied seedings for the 2017 French Open using four different periods of match history: 12, 9, 6, and 3 months. The chart below shows the results for each player that made it into the top 32 in at least one period (you can double click on an individual player to isolate her trend).

If player ability was constant over the past year, we would expect each player to have a flat line. But the lines we actually observe are much less stable. In fact, the median change per player from one period to the next was 4 seeds. Only 8 players showed good consistency in their performance level regardless of match history considered, including Elina Svitolina, Kristina Mladenovic, Simona Halep, and Venus Williams.

The table below shows the 32 French Open seeds that would have been selected by the surface-weighted Elo for each of the four different time periods. The final column shows the actual seeds for contrast. Players who are still in the draw at the time of this writing are highlighted in green.

Interestingly, the Elo-based seedings give a dramatically different order than the official seeds, no matter the time period considered. This is especially notable for the top seed, with none of the Elo approaches putting Kerber in the top spot but three of the four selecting Elina Svitolina. Considering the top 8 seeds, the 3-month Elo is perhaps performing the best at this stage, with 6 of 8 still in the draw and one of those two being Laura Siegemund who was unable to compete due to injury. The slight edge of a method that ignores all match results longer than 3 months ago is another indicator of the volatility of performance on the women’s tour this year.

Seed 3 months 6 months 9 months 12 months Official Seeds
1 Kristina Mladenovic Elina Svitolina Elina Svitolina Elina Svitolina Angelique Kerber
2 Elina Svitolina Kristina Mladenovic Caroline Wozniacki Simona Halep Karolina Pliskova
3 Simona Halep Mona Barthel Kristina Mladenovic Kristina Mladenovic Simona Halep
4 Laura Siegemund Simona Halep Johanna Konta Johanna Konta Garbine Muguruza
5 Anastasija Sevastova Caroline Wozniacki Svetlana Kuznetsova Caroline Wozniacki Elina Svitolina
6 Svetlana Kuznetsova Svetlana Kuznetsova Simona Halep Svetlana Kuznetsova Dominka Cibulkova
7 Caroline Wozniacki Johanna Konta Karolina Pliskova Venus Williams Johanna Konta
8 Anastasia Pavlyuchenkova Venus Williams Anastasia Pavlyuchenkova Karolina Pliskova Svetlana Kuznetsova
9 Mona Barthel Anastasia Pavlyuchenkova Venus Williams Anastasia Pavlyuchenkova Agnieszka Radwanska
10 Francesca Schiavone Karolina Pliskova Mona Barthel Garbine Muguruza Venus Williams
11 Venus Williams Anastasija Sevastova Anastasija Sevastova Angelique Kerber Caroline Wozniacki
12 Jelena Ostapenko Laura Siegemund Angelique Kerber Anastasija Sevastova Madison Keys
13 Kiki Bertens Mirjana Lucic Daria Gavrilova Laura Siegemund Kristina Mladenovic
14 Mirjana Lucic Jelena Ostapenko Laura Siegemund Agnieszka Radwanska Elena Vesnina
15 Lucie Safarova Lucie Safarova Catherine Cartan Bellis Mirjana Lucic Petra Kvitova
16 Lara Arruabarrena Garbine Muguruza Lauren Davis Mona Barthel Anastasia Pavlyuchenkova
17 Johanna Konta Sorana Cirstea Kristyna Pliskova Daria Gavrilova Anastasija Sevastova
18 Timea Bacsinszky Daria Gavrilova Garbine Muguruza Dominika Cibulkova Kiki Bertens
19 Karolina Pliskova Catherine Cartan Bellis Lesia Tsurenko Lauren Davis Coco Vandeweghe
20 Qiang Wang Francesca Schiavone Mirjana Lucic Catherine Cartan Bellis Barbora Strycova
21 Darya Kasatkina Ons Jabeur Agnieszka Radwanska Madison Keys Carla Suarez Navarro
22 Sorana Cirstea Coco Vandeweghe Jelena Ostapenko Timea Bacsinszky Mirjana Lucic-Baroni
23 Anett Kontaveit Lauren Davis Dominika Cibulkova Kiki Bertens Sam Stosur
24 Daria Gavrilova Timea Bacsinszky Lara Arruabarrena Katerina Siniakova Daria Gavrilova
25 Maria Sharapova Darya Kasatkina Lucie Safarova Darya Kasatkina Lauren Davis
26 Angelique Kerber Qiang Wang Sorana Cirstea Coco Vandeweghe Darya Kasatkina
27 Lauren Davis Lara Arruabarrena Katerina Siniakova Sorana Cirstea Yulia Putintseva
28 Shelby Rogers Angelique Kerber Francesca Schiavone Lesia Tsurenko Carloine Garcia
29 Sara Errani Kiki Bertens Darya Kasatkina Barbora Zahlavova Strycova Ana Konjuh
30 Catherine Cartan Bellis Barbora Zahlavova Strycova Julia Goerges Lucie Safarova Timea Bacsinszky
31 Irina Camelia Begu Shelby Rogers Barbora Zahlavova Strycova Julia Goerges Roberta Vinci
32 Julia Goerges Bethanie Mattek Sands Timea Bacsinszky Kristyna Pliskova Shuai Zhang
Stephanie Kovalchik avatar
About Stephanie Kovalchik
Tennis Data Scientist at the Game Insight Group at Tennis Australia and researcher at the Institute of Sport Exercise and Active Living at Victoria University.
comments powered by Disqus