Assessing Over & Under Performing ATP Players
Having a strong interest in sports analytics, reading this article which focuses on the success of FC Midtjylland and Brentford FC – https://decorrespondent.nl/2607/How-data-not-humans-run-this-Danish-football-club/230219386155-d2948861- naturally piqued my interest. These two clubs – both owned by millionaire hedge fund manager and sports bettor Matthew Benham – use data analytics to assess performance in a number of areas, and one way that they use these analytics is to look for undervalued players in the transfer market.
When I read this article, my first thought was how I could apply this to the Tennis markets. The Football transfer market and the Tennis betting/trading markets don’t make an easy comparison, but I decided that the main area it could be compared would be to look for undervalued players.
The easiest way of doing this would be to look for players whose results don’t bear their overall statistics justice. Perhaps these players have lost a number of tight sets or matches, or have under-performed in tiebreaks. In short, they’ve not excelled at key points. There may be reasons for this, such as poor mental strength, or a serve they cannot rely on to bail them out of trouble – my new Game State spreadsheets show that some players have a big issue when a set and break up, for example. However, variance could also play a big part in this and with Tennis a sport where fine margins often dictate the results of sets and matches, variance could be a major factor.
However, analysis will be able to indicate why this underperformance exists, and how we may be able to profit from this information.
I decided to use several metrics in an attempt to work out under and over performing players.
1 – The break point clutch score of that player (contained for each match daily in the ATP/WTA Spreadsheets) which measures how a player performs in break points, compared to their expectation from service and return points won %. A score of worse than -3% combined on serve and return (their break point save/won % is combined worse than -3% than their expectation based on service and return points won %) would indicate a player had performed very badly at break points.
2 – The tiebreak record of that player – also now available daily in the ATP/WTA Spreadsheets. The Game State Spreadsheets’ Tiebreak analysis illustrates tiebreaks are far from random, but shows that it is quite rare to generate a scenario where a player’s expectation to win a tiebreak is below 40%, and below 35% is incredibly unlikely. Therefore a player with a tiebreak win percentage of below 40% is under performing in tiebreaks, and below 35% is severely underperforming.
All things considered, players with a high clutch score and high tiebreak win percentage are highly likely to have a high match and set win percentage than their service hold/break opponent percentage would indicate. These players are over performing based on their ability. Conversely, those players with low clutch scores and tiebreak win percentages are failing to win key points and are likely to be better players than their overall match and set win percentage would indicate.
The table below illustrates the 12 month records from 26/3/14 to 26/3/15 for current top 100 players with a high or low Clutch Score in the 2013 season, where Clutch Score was greater than 3 or worse than -3, with a £100 hypothetical bet applied using Pinnacle Sports’ closing prices:-
| Player | 2013 Clutch Score | Matches | 12 Month P/L | 12 Month ROI |
| Mannarino |
14.1 |
88 |
-127 |
-1.4 |
| Sock |
12.8 |
52 |
976 |
18.8 |
| Haase |
7.3 |
62 |
406 |
6.5 |
| Klizan |
7.2 |
55 |
1711 |
31.1 |
| Tsonga |
6.9 |
38 |
547 |
14.4 |
| Istomin |
6.2 |
56 |
-535 |
-9.6 |
| Gulbis |
6.2 |
48 |
-838 |
-17.5 |
| Dodig |
5.1 |
38 |
-638 |
-16.8 |
| Youzhny |
4 |
41 |
-1061 |
-25.9 |
| Delbonis |
3.6 |
48 |
-1459 |
-30.4 |
| Almagro |
3.4 |
26 |
332 |
12.8 |
| Mayer L |
3.3 |
59 |
833 |
14.1 |
| Kohlschreiber |
3.2 |
57 |
-1100 |
-19.3 |
| Bautista-Agut |
3.2 |
62 |
168 |
2.7 |
| Dimitrov |
-3.4 |
63 |
-194 |
-3.1 |
| Gasquet |
-3.5 |
47 |
119 |
2.5 |
| Stepanek |
-3.7 |
31 |
15 |
0.5 |
| Pospisil |
-3.7 |
52 |
-499 |
-9.6 |
| Matosevic |
-3.8 |
56 |
242 |
4.3 |
| Giraldo |
-3.9 |
61 |
-25 |
-0.4 |
| Raonic |
-4 |
79 |
101 |
1.3 |
| Verdasco |
-4.3 |
53 |
-189 |
-3.6 |
| Lu |
-4.4 |
52 |
455 |
8.8 |
| Benneteau |
-4.4 |
44 |
175 |
4.0 |
| Kuznetsov Andrey |
-4.8 |
60 |
637 |
10.6 |
| Karlovic |
-5 |
63 |
2855 |
45.3 |
| Monaco |
-5.4 |
48 |
63 |
1.3 |
| Troicki |
-5.6 |
77 |
959 |
12.5 |
| Dolgopolov |
-5.7 |
37 |
-864 |
-23.4 |
| Ramos |
-5.8 |
83 |
-63 |
-0.8 |
| Querrey |
-5.9 |
66 |
-233 |
-3.5 |
| Djokovic |
-5.9 |
76 |
120 |
1.6 |
| Lorenzi |
-6.4 |
65 |
-937 |
-14.4 |
| Bellucci |
-6.5 |
59 |
-618 |
-10.5 |
| Stakhovsky |
-7.7 |
74 |
168 |
2.3 |
| Ferrer |
-7.7 |
75 |
713 |
9.5 |
Following this, we can then look at the data for players with a high or low clutch score:-
| Scenario | Matches | P/L | ROI |
| Clutch Score >3 |
730 |
-785 |
-1.08 |
| Clutch Score <-3 |
1321 |
3000 |
2.27 |
Players with a high clutch score recorded a return on investment of -1.08%, compared to 2.27% for those with a low clutch score. These figures illustrate that the players with a low clutch score are marginally under-rated by the market and they are open to some, albeit slight, improvement.
Moving on from this, we can look at the 12 month records of players who had over 65% or below 35% tiebreak win percentages in the 2013 season, with a £100 hypothetical bet applied using Pinnacle Sports’ closing prices:-
| Player | 2013 Tiebreak Win % | Matches | 12 Month P/L | 12 Month ROI |
| Bautista-Agut |
74 |
62 |
168 |
2.7 |
| Cilic |
71 |
50 |
15 |
0.3 |
| Lacko |
71 |
61 |
-613 |
-10.0 |
| Delbonis |
69 |
48 |
-1459 |
-30.4 |
| Djokovic |
68 |
76 |
120 |
1.6 |
| Nadal |
68 |
54 |
-837 |
-15.5 |
| Isner |
68 |
58 |
-994 |
-17.1 |
| Murray |
67 |
83 |
-268 |
-3.2 |
| Rosol |
33 |
65 |
33 |
0.5 |
| Monaco |
33 |
48 |
63 |
1.3 |
| Smyczek |
33 |
68 |
-110 |
-1.6 |
| Baghdatis |
32 |
51 |
790 |
15.5 |
| Giraldo |
30 |
61 |
-25 |
-0.4 |
| Andujar |
29 |
46 |
-127 |
-2.8 |
| Stepanek |
27 |
31 |
15 |
0.5 |
| Mannarino |
22 |
88 |
-127 |
-1.4 |
| Kavcic |
22 |
54 |
146 |
2.7 |
The summary of these percentages was much more stark:-
| Scenario | Matches | P/L | ROI |
| Tiebreak Win % > 65% |
492 |
-3868 |
-7.86 |
| Tiebreak Win % < 35% |
512 |
658 |
1.29 |
Players with a high tiebreak win percentage in the 2013 season returned very poor 12 month figures of -7.86% return on investment. A slight profit of 1.29% was recorded from backing players with a tiebreak win percentage below 35%.
From this data we can make the assertions that tiebreaks are more random than many people think, and opposing players with a strong tiebreak record is likely to yield dividends, as the market gives these players too much credit for winning a tight match.
The following players currently have a 12 month tiebreak win percentage of 65% or above:-
Milos Raonic
Stan Wawrinka
Gilles Muller
Gael Monfils
Marin Cilic
Kei Nishikori
Roger Federer
Richard Gasquet
Opposing these players, and Tour leaders Raonic, Wawrinka and Muller in particular, is a definite pre-match betting angle worth considering. It is also possible that these players may be over-rated by the market in-play, when a tiebreak occurs.
It is hugely important to delve beyond the win percentages of a player and try to understand how a player has won the match. It is very possible that a player that wins a tight match will win less points or games than their opponent, and win having had fewer break points than their opponent. Does this make them better on the day than the player they just beat? I’d argue that it doesn’t…
Many Thanks to Dan Weston who was the guest poster for this article, visit Dan’s excellent Tennis ratings site for lots of useful Tennis stats.
Category: Tennis, Trading strategies
