The ratings still show him to be the best in the world, writes Jack Houghton, so how do you deal with the return of Djokovic in the Australian Open?
"A weakness of ratings systems is that they only consider match data, ignoring other factors that might affect a player's performance. Going into the Australian Open, this causes us a problem..."
Most punters suffer from recency bias: a tendency to overvalue new information over old.
In tennis, this can be costly. A few early exits from tournaments is often enough for punters to consider a player to be in terminal decline. They rarely are, though. Top-flight men's tennis sees numerous encounters between closely-matched players, and this makes it statistically certain that some of them will, from time to time, experience losing streaks. Even matches where there is a large discrepancy in the known ability of players will see upsets.
When assessing the best bet for an upcoming tournament, then, giving too much credence to what happened in the last tournament isn't always wise: taking a longer view is usually best.
Take Roger Federer's surprise second-round exit from Wimbledon in 2013. Many saw it as the first stage of his inexorable collapse. Here was a player who had lost the glow of invincibility in recent seasons: Nadal frequently bettered him; he was no longer injury free; and he no longer exhibited that ruthless reliability when finishing off inferior players. His form, like the form of other tennis players, was assumed by some to be binary: he was either at his best or he was nothing at all. Federer was finished.
Except he wasn't, of course.
One of the advantages of using a computer-driven ratings system - like the Elo systems that most profitable punters utilise in some form - is that they don't overreact to recent results. Sure, if a player starts to lose more regularly, and against lesser opponents, their rating does decline, but more gradually. So, in 2013, whilst it was clear that Federer wasn't at his 2007 best, his Elo rating never ranked him lower than 6th-best in the world, explaining why, when recovered from injury, he resumed his usual practice of reaching the final stages of numerous tournaments with relative ease. Federer might no longer have been dominant, but he still won lots of matches.
A weakness of these kinds of ratings systems, though, is that they only consider match data, ignoring other factors that might affect a player's performance. Going into the Australian Open, this causes us a problem.
According to my ratings, Novak Djokovic is the best player in the world, holding between a 30- and 70-point advantage over Federer, depending on whether I examine the overall ratings, or only look at hard-court performances. This suggests that, in a head-to-head, Djokovic would have between a 54% and 60% chance of winning a match against Federer. But, if the two were to meet, who would take [1.85] (54%) on Djokovic?
The second-half of his 2016 season was marred by rumours of personal issues and disagreements with his coaching team, and he all-but sat out of 2017 with injury concerns. Reports suggest he has had only six weeks of court time in the run-up to the Australian Open, culminating in two exhibition matches on Wednesday, one of which he lost against now-retired Lleyton Hewitt.
When I plug my raw ratings into a simulation of the Australian Open, it seems Djokovic, with a relatively easy draw until a likely meeting with Federer in the semi-final, should be favourite, at around [4.00]. It's clear, though, that his Elo rating, only 300 points shy of its peak, does not accurately reflect the doubts around Djokovic's form.
What becomes more interesting is re-running the simulation with a reduced rating for Djokovic. To arrive at a new rating for Djokovic, I settled on a novel - and perhaps flawed - method.
Given what we know about Djokovic's form (he's doubtful over his fitness and was unimpressive in his exhibition matches), I considered what odds would I set were he to play Federer in the first-round? I settled on Federer being a [1.20]-shot. Working backwards, this translates to saying that Djokovic's Elo rating should be 280 points lower than Federer's. This marks Djokovic out as the 12th-best player in the drawer and significantly reduces his chances of winning the tournament, according to my simulation, to around [60.00] - far longer odds than the [8.80] available.
There are obviously questions to be asked about such a rudimentary adjustment of a player's rating, and others may set different odds in a supposed match-up between Federer and Djokovic, but in the absence of a better approach, I'm convinced at least that the odds for Novak Djokovic winning are too short based on what we know.
Conversely, Roger Federer ([3.00]) and Nick Kyrgios ([19.00]) are both value bets. With the absence of Andy Murray and perennial doubts around Nadal's injuries, supporting both gives us a player to cheer on both sides of the draw.
Back Roger Federer at [3.00] and Nick Kyrgios at [19.00] to win Australian Open.