Australian Open 2018 Glicko Preview

Martin Ingram – Data Scientist

Martin is passionate about good code and gaining insights from data with a particular interest in sports analytics. He completed a Bachelor Of Arts, Natural Sciences (Physical) at Cambridge before completing his Masters of Science, Computing Science.


Australian Open Glicko Preview

The Australian Open 2018 is around the corner and it promises to be an intriguing event. Many top players on both the ATP and WTA are planning to return from long absences, with Novak Djokovic and Andy Murray heading the list on the ATP. And it’s confirmed that Serena Williams is out from the WTA side.

In 2017, Roger Federer surprised the tennis world by coming back from an injury break and winning his first Grand Slam title in five years. Will 2018 see more success for Rafael Nadal and Roger Federer, or will Novak Djokovic and Andy Murray return as strong as Federer did last year? In this article, I use the Glicko rating system, a generalisation of Elo, to pick the likely winners on both the ATP and WTA.

Long absences of players complicate prediction, because it is not clear whether players will return to play at the strength they left. Sometimes they do, like Federer did last year. Other times, players have more mixed results, like Sharapova did in her comeback last year. The popular rating system Elo does not take this uncertainty into account, instead predicting that players play at the same skill after a year-long break as after a two-week rest, which seems questionable. Motivated by this shortcoming, I introduce Glicko in this article, a generalisation of the Elo ratings which deals with injury breaks in a more natural way than Elo.


Glicko is a rating system developed by Mark Glickman, Senior Lecturer of Statistics at Harvard University. Popular for chess, it improves on Elo by not only modelling a player’s skill, but also the uncertainty about that skill.

Glicko breaks time down into periods – I use weeks in this case. At each period, a player has a mean skill  and an uncertainty  about that skill. Just like for Elo, players start out with a mean of =1500. The optimal value of the initial uncertainty can be found by optimising Glicko’s predictive performance, and turns out to be  on the WTA, and  on the ATP, indicating that the initial spread of skills is somewhat larger on the WTA.

Within a period, each player’s Glicko ratings are updated according to their match results: upwards if players perform better than expected, and downwards if they play worse. The more uncertain a player’s skill, the larger the update. Each observation reduces the uncertainty about a player’s skill.

Between periods, uncertainty is added, since players may improve or worsen over time. This allows Glicko to handle injury breaks better than Elo: if a player does not play, their mean remains the same, but the uncertainty about their skill grows. When they finally play again, Glicko is prepared to make larger changes to their mean to be able to adjust to a potential drop in performance.

The amount of uncertainty added between periods is another parameter that can be found by optimising Glicko’s predictive accuracy. It turns out to be  on the WTA, and  on the ATP, indicating that players on the WTA tour tend to be more variable from one week to the next than players on the ATP.

Given two Glicko ratings  for player 1 and  for player 2, player 1’s win probability can be calculated as:


and .

For more information on Glicko, please see Mark Glickman’s explanation or the original paper.

ATP Glicko Predictions

Without further ado, here are the top 12 favourites on the ATP:

Player  μ  σ
(1) Roger Federer 2109 68
(2) Novak Djokovic 2076 81
(3) Rafael Nadal 2018 61
(4) Andy Murray 1993 78
(5) Juan Martin Del Potro 1901 59
(6) Kei Nishikori 1885 71
(7) Milos Raonic 1870 70
(8) Stan Wawrinka 1852 72
(9) Grigor Dimitrov 1849 53
(10) Nick Kyrgios 1842 67
(11) Alexander Zverev 1830 55
(12) Jo-Wilfried Tsonga 1813 63


According to Glicko, Roger Federer, two-time Grand Slam champion in 2017, is currently the world’s strongest player and favourite to win the Australian Open. He posted strong results towards the end of 2017, winning Shanghai and Basel, before ending the season with an unexpected loss to David Goffin at the ATP World Tour Finals.

Despite his mixed results in 2017, Novak Djokovic comes in second, but note that because he cut his 2017 season short after Wimbledon, the uncertainty about his skill is the highest of all 12 players in the list, indicating that Glicko is somewhat sceptical of whether he can live up to his high rating. He has recently pulled out of the Abu Dhabi exhibition with an elbow injury, casting further doubt on his ability to come back strongly.

World number 1 Rafael Nadal follows as the third favourite. His gap to Federer is not huge: he would have a probability of 37% of beating him according to Glicko. However, he has recently pulled out of Brisbane with knee issues, suggesting he may not be fully fit for the event. Andy Murray completes the top 4, and once again, his long break causes Glicko to have a large uncertainty about his skill.

Behind these top 4, several players follow who are around 200-250 points from Federer. Juan Martin Del Potro has a 23% chance of repeating his victory against Federer at last year’s US Open if they were to play again. He finished the last year quite strongly and could be a force to be reckoned with.

More doubts surround the number six, seven and eighth-ranked players, Nishikori, Raonic and Wawrinka. Once again, their uncertainties are quite high at over 70. Raonic has only played two matches since Washington in August 2017; Wawrinka last played when he lost his first round in Wimbledon last July.

Glicko is much more certain about Grigor Dimitrov, who finished last year very strongly with a victory in the ATP World Tour Finals. He could be a player to look out for, but if Federer, Djokovic or Nadal play at full strength, he would only have a 18%, 21% and 27% chance against them, respectively.

Two young players, Kyrgios and Zverev, follow closely behind. Kyrgios had a season with ups and downs in 2017, but had some highlights towards the end, reaching the finals of Beijing and losing only to Rafael Nadal. Zverev had a strong season overall, winning two Master’s tournaments, but appeared to run out of steam by the end, losing six of his last nine matches. Zverev has struggled at Grand Slams in the past, so it will be fascinating to see whether he can come back rejuvenated and finally play strongly at a major.

Finally, Jo-Wilfried Tsonga comes in at number 12. A solid player, he finished 2017 quite strongly with victories in Antwerp and Vienna.

Overall, the ATP’s favourites are riddled with players coming back from injury, making it hard to make definitive predictions. Roger Federer and Rafael Nadal (if healthy) are the favourites, but while sceptical due to their long absences, Glicko still scores Murray and Djokovic highly. It will be fascinating to see whether they can come back strongly.

ATP Winners and Losers

Although still ranked highly, Andy Murray and Novak Djokovic are by far the biggest losers among players ranked in the top 32 by Glicko compared to last year. They have dropped 160 and 116 points, respectively, since January 2017. To put this into perspective: current Andy Murray would only have a 29% chance of beating Andy Murray one year ago according to Glicko. Kei Nishikori, Tomas Berdych and Gael Monfils also did not have a great year, dropping around 60 points.

The biggest winner on the ATP is Adrian Mannarino, gaining 105 points, followed by Rafael Nadal (+97), Diego Sebastian Schwartzman (+92), Alexander Zverev (+87), Roger Federer (+83) and Sam Querrey (+81).

WTA Glicko Predictions

Here are the top 12 favourites on the WTA according to hard-court Glicko:

Player μ σ
(1) Serena Williams 2077 117
(2) Victoria Azarenka 1942 114
(3) Maria Sharapova 1914 93
(4) Simona Halep 1882 67
(5) Caroline Wozniacki 1873 63
(6) Elina Svitolina 1868 68
(7) Garbine Muguruza 1851 66
(8) Venus Williams 1834 68
(9) Karolina Pliskova 1824 65
(10) Madison Keys 1803 76
(11) Caroline Garcia 1799 60
(12) Sloane Stephens 1794 77


High uncertainty is even more striking on the WTA than on the ATP. The three highest-ranked players are Williams, Azarenka and Sharapova, but all three have very large terms. This makes sense. Williams is out, Azarenka played only six matches in all of 2017 after a year’s absence, and Sharapova is not too far into the comeback she started in 2017. Azarenka is estimated to have a 59% chance of beating Simona Halep, but given her long break, this is very questionable.

At 93, Sharapova’s uncertainty is high but somewhat lower than Azarenka’s. She played 21 matches in 2017 and most recently had a good result, winning Tianjin, before losing in the first round of Moscow. Although she had some success last year, it is not yet clear whether Sharapova can play as well as she did before her doping ban.

After these three high-uncertainty players, players with much lower values of  follow. Halep, Wozniacki and Svitolina are ranked within 14 points of each other, making them virtually indistinguishable according to Glicko. Halep had a mixed end to 2017, reaching the finals of Beijing before losing in the group stages of the WTA Championships. Wozniacki, on the other hand, finished very strongly, beating both Halep and Svitolina at the same event to win the WTA Championships. Svitolina also had more of a mixed end to the year, losing in the group stages of the Championships.

Wimbledon winner Garbine Muguruza follows closely behind, and in contrast to Wozniacki, Svitolina and Halep, she has won Grand Slams in the past, which could give her a mental edge. The same goes for Venus Williams, who played particularly strongly in the big events last year, reaching the finals of the Australian Open, Wimbledon and the WTA Championships. It is hard to count these players out. With scores of 1851 and 1834, they are close to Halep and would have a 46% and 43% chance of beating her, respectively.

Madison Keys, Caroline Garcia and Sloane Stephens complete the top 12 favourites. Madison Keys reached the US Open final last September but has only played one match since; Sloane Stephens beat her in that final but struggled after that, losing in the first round of Wuhan and Beijing. Caroline Garcia, on the other hand, had a very strong finish to the year, winning both Wuhan and Beijing before losing in the semi-finals of the WTA Championships. If she can continue her hot streak, she could be worth looking out for.

WTA Winners and Losers

On the WTA, the biggest loser by far is Angelique Kerber. After a terrible 2017, she dropped almost 200 points compared to January 2017, making her collapse even more dramatic than Andy Murray and Novak Djokovic’s on the ATP. Agnieszka Radwanska (-106), Maria Sharapova (-90), Petra Kvitova (-76), Svetlana Kuznetsova (-66) and Victoria Azarenka (-64) also struggled somewhat.

On the other hand, Aussie rising star Ash Barty improved hugely, increasing her Glicko rating by a whopping 223 points. She barely missed the top 12, coming in at number 13, and is worth looking out for at the Australian Open. Jelena Ostapenko (+177), Shuai Peng (+171), Magdalena Rybarikova (+163), Caroline Garcia (+152) and Julia Goerges (+132) also improved hugely. Sloane Stephens (+121) and Anastasija Sevastova (+108) round out the winners on the WTA.


Long injury breaks make it hard to predict this year’s Australian Open. It will be fascinating to see whether Novak Djokovic and Andy Murray can come back at full strength. If they do not, Federer – and Nadal, assuming he recovers from his knee issues — are strong favourites on the ATP.

Much like 2017, the WTA field is wide open, with many of the favourites ranked very closely together. Halep, Svitolina, Wozniacki and Muguruza are the strongest contenders, and Venus Williams is not far behind.

Betting Strategy


Back on Betfair BACK Roger Federer at 3.30 or bigger for 2 units.


Back on Betfair BACK Simona Halep at 9 or bigger for 1 unit.


Related Articles

Tennis Modelling: ELO

Tennis Modelling: Introduction

Related Articles