The date is January 27, 2017, and it is the men’s semi-final at the Australian Open. Grigor Dimitrov is due to take on Rafael Nadal, with a place in the final against Roger Federer at stake.
Tennis media, tennis fans and tennis bettors alike are starting to believe the hype - that Dimitrov, at almost 26 years of age, is finally joining the elite and finally living up to his reputation as ‘Baby Fed’.
Break point overperformance
Numbers, however, tell a different story. Dimitrov, in 10 consecutive wins from 10 matches in 2017 to this point, had overperformed by 9.7% above expectation for saving break points on his serve (based on service points won expectation) and 3.4% above expectation for converting break points on return (based on return points won expectation).
Maintaining such rates in the long term is virtually impossible - and it was as good as certain that Dimitrov would mean revert eventually. Nadal won in five sets that day and yet, following that defeat, Dimitrov took the title in Sofia in his next event, dropping just one set en route to the title.
Maintaining such rates in the long term is virtually impossible - and it was as good as certain that Dimitrov would mean revert eventually.
The hype continued; not surprising given his pricing of 1.41 or shorter in all four of his matches in Bulgaria - find out how to take advantage of tennis handicap betting when a player is such a heavy favourite.
Subsequently, Dimitrov won just five of 12 matches between Sofia and the French Open, including five losses as a 1.50 favourite or below, and the hype soon quietened. Mean reversion, as expected, kicked in, and he levelled out to being no better than he was at the end of the 2016 season - not good enough to be top 10, but still a solid top-20 player.
Variance not 'form'
Such fluctuations are frequently passed on as 'form' when they should be more accurately described as variance. It is very difficult indeed for a player to maintain overperformance at break points, both on serve and return, over the course of numerous seasons - if you know how to bet on tennis, you will know how important break points can be.
With this in mind, analysis was prepared to assess whether the market was biased towards players who had overperformed on break points, similar to Dimitrov in the Australian Open match previously referred to.
Firstly, it’s worth looking at serve-orientated players on the ATP Tour to see whether they saved more break points than expected. Using a simple metric of (service hold % - break opponent %), we can calculate which players are the most serve-orientated performers in the top 100.
For this sample, data was used for the players in the top 100 rankings at the end of the 2016 season (minimum ten main tour matches in 2016 were required for a player to qualify in the sample).
Top 10 most serve-orientated players:
Player
|
End 2016 Rank
|
2016 Hold %
|
2016 Break %
|
2016 Hold/Break difference %
|
Karlovic
|
20
|
92.9
|
7.3
|
85.6
|
Isner
|
19
|
93.4
|
10.4
|
83.0
|
Muller
|
34
|
88.7
|
13.0
|
75.7
|
Raonic
|
3
|
90.5
|
18.3
|
72.2
|
Johnson
|
33
|
85.1
|
13.3
|
71.8
|
Zeballos
|
71
|
81.4
|
11.0
|
70.4
|
Querrey
|
31
|
85.6
|
15.6
|
70.0
|
Kyrgios
|
13
|
88.7
|
19.6
|
69.1
|
Cuevas
|
22
|
85.5
|
17.9
|
67.6
|
Tsonga
|
12
|
87.7
|
20.4
|
67.3
|
The majority of the names on this list should not be a surprise to those familiar with the tennis world, with both Ivo Karlovic and John Isner by some distance the most serve-orientated players on the ATP Tour.
With such high service hold percentages, derived from high service points won numbers, it is obvious that these players will also have high break point save percentages. As the next table illustrates, these players managed to save break points far more often than the average 60.7% ATP mean percentage for main draw matches in 2016:
2016 Break Point Performance
Player
|
2016 Break Point Save %
|
2016 Overall BP Over/Underperformance %
|
Karlovic
|
72.7
|
4.3
|
Isner
|
69.3
|
-2.9
|
Muller
|
65.0
|
-3.1
|
Raonic
|
69.4
|
-2.3
|
Johnson
|
64.4
|
-2.7
|
Zeballos
|
64.8
|
9.3
|
Querrey
|
64.9
|
-0.1
|
Kyrgios
|
69.4
|
2.4
|
Cuevas
|
65.2
|
0.9
|
Tsonga
|
67.4
|
-0.6
|
The average ATP player in main draw matches saves 2.8% fewer break point chances on their serve than they win service points, while they convert 2.8% more break point chances on return than their return points won percentage. We can use these numbers to establish break point over/underperformance.
When this is factored in, six of the top ten most serve-orientated players underperformed on break points, with Gilles Muller the biggest culprit on key points and Horacio Zeballos, by some distance, the biggest overperformer. The data is therefore inconclusive as to whether serve-orientated players over-performed more on key points than average.
Variance being overrated by the market
Digging deeper, the table below shows the top ten players who overperformed on break points in 2016. We can use this and data so far from main tour matches in 2017 to assess whether the market overrated these players after variance favoured them in 2016:
Top ten players who overperformed on break points in 2016
Player
|
End 2016 Rank
|
2016 Overall BP Over/Underperformance %
|
Brown
|
72
|
13.9
|
Berlocq
|
95
|
10.5
|
Zeballos
|
71
|
9.3
|
Evans
|
66
|
8.1
|
Monteiro
|
82
|
7.5
|
Klizan
|
35
|
6.9
|
Marchenko
|
74
|
6.3
|
Khachanov
|
53
|
4.7
|
Harrison
|
90
|
4.5
|
Herbert
|
78
|
4.4
|
Using a hypothetical £100 flat stake, backing these players blind in 2017 would generate the following returns:
Betting on players who overperformed on break points
Player
|
Matches
|
Wins
|
Win %
|
P/L
|
ROI %
|
Brown
|
22
|
8
|
36.36
|
-98
|
-4.45
|
Berlocq
|
24
|
9
|
37.50
|
-642
|
-26.75
|
Zeballos
|
32
|
15
|
46.88
|
163
|
5.09
|
Evans
|
17
|
9
|
52.94
|
645
|
37.94
|
Monteiro
|
23
|
7
|
30.43
|
-797
|
-34.65
|
Klizan
|
20
|
8
|
40.00
|
-309
|
-15.45
|
Marchenko
|
6
|
2
|
33.33
|
-147
|
-24.50
|
Khachanov
|
36
|
17
|
47.22
|
-272
|
-7.56
|
Harrison
|
28
|
14
|
50.00
|
-503
|
-17.96
|
Herbert
|
18
|
6
|
33.33
|
-214
|
-11.89
|
Overall
|
226
|
95
|
42.04
|
-2174
|
-9.62
|
As can be seen, blind-backing these players who overperformed the most on break points in 2016 would have led to disastrous returns so far this year, returning -9.62% ROI from 226 matches.
In effect, the market has overvalued the ‘form’ and ability level of these players due to their key point overperformance. With the exception of Dan Evans - now serving a ban – and, to an extent, Horacio Zeballos, all of the other ten players have mean reverted in 2017.
With such poor results blind-backing these players, taking the approach of opposing them, in anticipation of mean reversion, would have yielded positive returns. It is also interesting to assess whether players who underperformed on break points the most in 2016 would yield positive returns in 2017, with the market underestimating their abilities:
Top ten players who underperformed on break points in 2016
Player
|
End 2016 Rank
|
2016 Overall BP Over/Underperformance
|
Sela
|
96
|
-10.4
|
Medvedev
|
99
|
-8.2
|
Del Potro
|
38
|
-7.2
|
Sousa
|
43
|
-7.1
|
Dimitrov
|
17
|
-7.0
|
Federer
|
16
|
-7.0
|
Coric
|
48
|
-6.9
|
Struff
|
63
|
-6.5
|
Almagro
|
44
|
-5.8
|
Mayer F
|
50
|
-5.7
|
Interestingly, we can see that Dimitrov, who overperformed so severely on break points early this year, actually underperformed significantly in 2016 overall on the same metric, aptly demonstrating how easy it is to disguise variance as ‘form.’
Mean reversion
Backing these players blindly was something of a mixed bag but, led by Roger Federer who has won 34 of 36 matches this season*, doing so did yield almost 6% return on investment from a hypothetical £100 flat stake from 274 matches:
Betting on players who underperformed on break points
Player
|
Matches
|
Wins
|
Win %
|
P/L
|
ROI %
|
Sela
|
16
|
5
|
31.25
|
66
|
4.13
|
Medvedev
|
32
|
19
|
59.38
|
690
|
21.56
|
Del Potro
|
25
|
14
|
56.00
|
-542
|
-21.68
|
Sousa
|
33
|
15
|
45.45
|
-653
|
-19.79
|
Dimitrov
|
40
|
26
|
65.00
|
-280
|
-7.00
|
Federer
|
36
|
34
|
94.44
|
2399
|
66.64
|
Coric
|
27
|
13
|
48.15
|
393
|
14.56
|
Struff
|
29
|
13
|
44.83
|
-638
|
-22.00
|
Almagro
|
17
|
8
|
47.06
|
-60
|
-3.53
|
Mayer F
|
19
|
8
|
42.11
|
204
|
10.74
|
Overall
|
274
|
155
|
56.57
|
1579
|
5.76
|
*Data from before the Rogers Cup semi-finals
To summarise, looking at the 2017 figures for players who over and underperformed on break points in 2016, there certainly is a strong case for asserting that mean reversion took place for both groups – and that the market has not accurately accounted for this likelihood.
Further research from previous years, to build up a bigger sample, may reward readers with an interesting angle to attack the market. Want more help with betting on tennis? Read the highlights from Pinnacle's tennis betting Discussion Day.