Since there is no tennis on the Lord's Day, it's a great opportunity to look at where things stand going into the second week of the gentlemen's singles draw at Wimbledon 2021. In particular, I was curious about what each of the 16 players' odds in terms of percentage are of lifting the trophy at the end of the fortnight.
Logically, every player has a 1 in 16 chance, which comes out to about a six percent chance of winning the title. However, not all players are created equally, nor is the draw, leaving some players with a better chance than others.
What I want to do here is explain my methodology for determining the percentage odds that each of the 16 players win the title. Then I will update those odds throughout the second week on my Twitter account (@jaredpinetennis). But before we get any farther, here are the odds for each player.
I developed this predictor just over the last two days, but some of the methodology is based on work I have done in other areas. I've never been taught the Elo rating system or Elo predicting methods, so this predictor is very amateur, but it is also entirely independent in its methodology as a result, which makes it unique.
The first part of developing this predictor was to come up with a rating for each player. Someone smarter than me could have figured out a way to rate all 128 players, but I could only fit 38 names on an excel sheet on my screen at once, so I only ranked the top 38 players in the current ATP rankings. That left three of the players in the round of 16 off the rating chart, so I simply listed them at replacement level.
Then I collected all results of those 38 players going back to Cincinnati 2021. Based on their results, each player received a rating between 16.2 (Djokovic) and -11 (Fognini) with a method that would take too long to explain here. The ratings that were produced correspond closely to the current ATP rankings.
The next step was the calculate what the odds of an upset are. Not all upsets are the same though. Zverev beating Tsitsipas is much more likely than Goffin beating Djokovic. The way I calculated the odds of an upset was to look at the frequency with which a player beat someone that they were four points better than over the last 12 months. The answer was 68% of the time. So if you are playing someone that you are four points better than, you have approximately a 68% chance of winning that match. Sample sizes were still a little small even with 12 months of data to work with, so I used a rolling average to smooth out some of the outliers. The result was that there is a 53% chance of beating someone that you are two points better than, and an 88% chance of beating someone you are 20 or more points better than.
I then used this method to calculate the odds that each player would win their next four matches. The difficulty starts with the second match. Nobody knows who they will face in the quarterfinals if they win in the fourth round. Therefore, it was necessary to determine the odds also that each player had of facing each opponent. So to determine the odds that each player has of reaching the quarterfinals, the formula is: odds of reaching the quarterfinals multiplied by the odds of beating the first possible opponent multiplied by the odds of that opponent reaching the quarterfinals plus the odds of beating the second possible opponent multiplied by the odds of that opponent reaching the quarterfinals. Remember your order of operations.
This gets even more complicated when there are four possible opponents in the semifinals and eight possible opponents in the final. Knowing how to right code would have made this much easier than doing all the calculations in an excel sheet. I really should have paid more attention in my computer science class, instead of developing an MVP calculator for the MLB while my professor was lecturing.
There are 32,768 different ways that the last week of Wimbledon can turn out. This methodology essentially considers all 32,768 possibilities, then it calculates the odds of each of these unique outcomes, and finally it adds up the odds of all the 2048 outcomes that have Djokovic as champion for a sum of a 33% chance that Djokovic will be the champion.
Before I finish writing, I do want to acknowledge a few weaknesses in the methodology in order to help anyone interesting to rightly interpret the data in the table. The first weakness is that in considering all the results since Cincinnati 2020, results on clay are over-represented, because Rome and Roland Garros have both been played twice in that span. That means players that excel on clay are overrated in this system. In particular, Sonego is slightly overrated and I think Shapovalov is underrated.
The other main weakness is that the rating system does not take into consideration whether a match was played at Wimbledon or a 250 on clay. Certainly a result from last week at Wimbledon should carry more weight for predicting futures matches at Wimbledon than a best-of-3 match played on clay back in October 2020.
The way that both of these areas of weakness in the methodology impact the final data is that it inflates the possibility of an upset. That's good news for Djokovic and Medvedev fans, who probably think their players have a better chance of winning than just 33 or 27 percent.
What will be interesting to watch now is to see how the odds go up with each passing match. As the list of possible champions drops from 16 to 8 over the course of Manic Monday, how much will that help each players' odds of winning the title? I'm especially curious to see how a result on the opposite side of the draw might affect a players' odds. Obviously, if Djokovic loses, everybody's odds will go up. But if Hurkacz loses to Medvedev, what impact if any will that have on Djokovic's odds of winning the title?
These are the kind of things I want to keep an eye on as I update the odds on Twitter throughout the next week.
No comments:
Post a Comment