Reality Check: Is the Wimbledon seeding system a good predictor?
Reality Check: Is the Wimbledon seeding system a good predictor?
(2 days later)
The seeding system for the men's singles at Wimbledon is different to the systems used at the other grand slam tournaments.
Wimbledon's seeding system for the men's singles has made an unusually big difference this year, as you can see from the lists above of the ATP rankings a week before the tournament and the Wimbledon rankings.
The French Open, US Open and Australian Open just take the ATP rankings, but Wimbledon works it out differently to favour players who are better on grass.
Wimbledon uses a system that favours grass-court specialists - taking the ATP ranking points, doubling the points earned at grass tournaments in the past year and adding on 75% of the points earned on grass the previous year. The other grand slams just use the ATP rankings.
The seedings are important because a highly seeded player cannot meet another highly seeded player until late in the tournament.
There are usually only two or three changes in the top eight seeds each year.
Wimbledon takes the points used to calculate a player's ATP ranking, doubles the points earned at grass tournaments in the past year and adds on 75% of the points earned on grass the previous year.
Over the last five years you could classify three of the changes to the top eight seeds as being "good" in that they make the seeding a better predictor of the outcome, and four of them as "bad" because they make it a worse predictor.
The ladies singles system is different in that a committee decides whether the WTA rankings need changing. No changes were made for 2017. A seeding committee was also used for the men's competition until 2002.
The highest profile example came in 2014, when Novak Djokevic was made number one seed at Wimbledon, despite being number two on the ATP rankings. He won the tournament while Rafa Nadal, who was demoted to the number two slot, was knocked out in the fourth round.
How much difference does it make?
On the other hand, in 2012 Tomas Berdych was promoted above David Ferrer in the seedings and was knocked out in the first round, while David Ferrer reached the quarter-finals.
If you compare the top eight seeds in the men's singles for the last five years with the players at the top of the ATP rankings one week before the tournament each year, there have only been two or three changes each year.
The difference has been marginal overall, but it also must be taken into account that changing seedings is partly a self-fulfilling policy, because a higher-seeded player is likely to get further in the tournament as a result of playing lower-ranked players.
Over that five year period you could classify five of the changes as being "good" in that they make the seeding a better predictor of the outcomes and three of them as "bad" because they make it a worse predictor.
Looking at how much difference the Wimbledon seeding system makes got the Reality Check team wondering about whether it had been a better predictor than seedings at other grand slams.
The highest profile example came in 2014, when Novak Djokovic was made number one seed at Wimbledon, despite being number two on the ATP rankings.
To compare the seedings with the outcomes for the top eight seeds in grand slams from 2012 to 2016, we allocated a numerical value for the stage at which a player was knocked out. For example, a player knocked out in the semi-finals gets a value of 3.5, because he could have come either third or fourth. Similarly, someone knocked out in the first round would get a value of 96.5.
Djokovic won the tournament while Rafael Nadal, who was demoted to the number two slot, was knocked out in the fourth round.
If the seeding system was perfect then adding up the outcomes for the top eight seeds in a single year would give a total of 36 (one + two + 3.5 + 3.5 + four lots of 6.5). In fact, the average number you get for the last five years at Wimbledon is 146. And actually, you also get 146 if you do the calculation with the ATP rankings instead of the Wimbledon seedings.
On the other hand, in 2012 Tomas Berdych was promoted above David Ferrer in the seedings and was knocked out in the first round while David Ferrer reached the quarter-finals.
But that is considerably higher than the figures of 106 at the US Open, 93 at the French Open and 89 at the Australian Open. It should be said that all of these numbers are pretty high. There is not a strong correlation between seeding and outcome.
It also must be taken into account that changing seedings is partly a self-fulfilling policy, because a higher-seeded player is likely to get further in the tournament as a result of playing lower-ranked players.
Nonetheless, it is much worse an indicator at Wimbledon, suggesting that Wimbledon has been a less predictable tournament over the past five years than the other grand slams.
Over the past five years, there have been four "bad" changes in the top eight seeds and three "good" changes, so the Wimbledon seeding method has been marginally worse on that measure than just using the ATP rankings, but there is not a lot in it.
Correction 10 July 2017: This report has been updated to include rankings for the 2017 tournament and to correct some outcomes from the analysis.
To find out whether it has been a better predictor than seedings at other grand slams Reality Check compared the seedings with the outcomes for the top eight seeds in grand slams from 2012 to 2016.
The outcomes allocate a numerical value for the stage at which a player is knocked out, for example, a player knocked out in the semi-finals gets a value of three or four because he could have come either third or fourth. Similarly, someone knocked out in in the first round would get a value of 96.5.
If the seeding system was perfect then adding up the outcomes for the top eight seeds in a single year would give a total of 36 (one + two + 3.5 + 3.5 + four lots of 6.5).
In fact, the average number you get for the last five years at Wimbledon is 146.
That is considerably higher than the figures of 106 at the US Open, 93 at the French Open and 89 at the Australian Open.
First of all, it should be said that all of these numbers are pretty high. There is not a strong correlation between seeding and outcome.
Nonetheless, it is much worse an indicator at Wimbledon. It would be tempting to conclude that was because the seeding system is different, but we have already established that the changes made only make a marginal difference.
That means that Wimbledon must just have been a less predictable tournament over the past five years than the other grand slams.
