Simpson's Paradox: Best and Worst Players

money_ball

Rookie
I recently read the following article:
So I decided to take a look at the stats myself. I queried all the matches since 1991 that had available data on "total winner points won" and "total loser points won", and looked up all the matches where the loser of the match won more points than the winner.

The following is the list of players who had at least 20 match wins where they won less total points than the loser:
Code:
Player                  W   L   Win %
------                  --  --  ------
John Isner              37  21  63.79%
Carlos Moya             33  15  68.75%
Tommy Robredo           33  14  70.21%
Feliciano Lopez         32  19  62.75%
Fabio Fognini           28  6   82.35%
Ivo Karlovic            26  26  50.00%
Goran Ivanisevic        25  19  56.82%
Marat Safin             24  17  58.54%
Fernando Gonzalez       24  8   75.00%
Mikhail Youzhny         23  18  56.10%
Jarkko Nieminen         23  15  60.53%
Andreas Seppi           23  15  60.53%
Tommy Haas              22  17  56.41%
Yevgeny Kafelnikov      21  13  61.76%
Gilles Muller           21  8   72.41%
Tomas Berdych           20  15  57.14%
Alex Corretja           20  14  58.82%
Rafael Nadal            20  14  58.82%
Jack Sock               20  10  66.67%
Guillermo Garcia-Lopez  20  7   74.07%
Greg Rusedski           19  19  50.00%
Gustavo Kuerten         19  17  52.78%
Gael Monfils            19  13  59.38%
Julien Benneteau        19  10  65.52%
Cedric Pioline          18  23  43.90%
Wayne Ferreira          18  23  43.90%
Jan Siemerink           18  18  50.00%
Mark Philippoussis      18  16  52.94%
Lleyton Hewitt          18  14  56.25%
Younes El Aynaoui       18  12  60.00%
Jo-Wilfried Tsonga      18  10  64.29%
Felix Mantilla          18  4   81.82%
Pete Sampras            17  20  45.95%
Fabrice Santoro         17  18  48.57%
Javier Sanchez          17  12  58.62%
Teymuraz Gabashvili     17  11  60.71%
Jurgen Melzer           17  9   65.38%
Andrei Medvedev         17  9   65.38%
Kei Nishikori           17  6   73.91%
Paul-Henri Mathieu      16  20  44.44%
Tim Henman              16  20  44.44%
David Ferrer            16  19  45.71%
James Blake             16  18  47.06%
Juan Carlos Ferrero     16  16  50.00%
Andre Agassi            16  15  51.61%
Francisco Clavet        16  15  51.61%
Marc Rosset             16  11  59.26%
Daniel Vacek            16  4   80.00%

The following is the list of players who had at least 16 match losses where they won more total points than the winner:
Code:
Player                  W   L   Win %
------                  --  --  ------
Roger Federer           8   32  20.00%
Ivo Karlovic            26  26  50.00%
Fernando Verdasco       15  26  36.59%
Richard Krajicek        14  26  35.00%
Todd Martin             11  26  29.73%
Cedric Pioline          18  23  43.90%
Wayne Ferreira          18  23  43.90%
Sam Querrey             10  23  30.30%
Juan Monaco             9   23  28.13%
Ivan Ljubicic           14  22  38.89%
Max Mirnyi              9   22  29.03%
Nikolay Davydenko       7   22  24.14%
John Isner              37  21  63.79%
Pete Sampras            17  20  45.95%
Paul-Henri Mathieu      16  20  44.44%
Tim Henman              16  20  44.44%
Nicolas Almagro         14  20  41.18%
Dominik Hrbaty          7   20  25.93%
Feliciano Lopez         32  19  62.75%
Goran Ivanisevic        25  19  56.82%
Greg Rusedski           19  19  50.00%
David Ferrer            16  19  45.71%
Philipp Kohlschreiber   15  19  44.12%
Thomas Enqvist          14  19  42.42%
Michael Chang           12  19  38.71%
Andrei Pavel            11  19  36.67%
Xavier Malisse          7   19  26.92%
Marin Cilic             7   19  26.92%
Mikhail Youzhny         23  18  56.10%
Jan Siemerink           18  18  50.00%
Fabrice Santoro         17  18  48.57%
James Blake             16  18  47.06%
Stan Wawrinka           15  18  45.45%
Nicolas Lapentti        15  18  45.45%
Thomas Johansson        14  18  43.75%
Richard Fromberg        10  18  35.71%
Sebastien Grosjean      8   18  30.77%
Marat Safin             24  17  58.54%
Tommy Haas              22  17  56.41%
Gustavo Kuerten         19  17  52.78%
Gilles Simon            15  17  46.88%
Andy Roddick            13  17  43.33%
Mark Philippoussis      18  16  52.94%
Juan Carlos Ferrero     16  16  50.00%
Albert Costa            13  16  44.83%
Richard Gasquet         10  16  38.46%
Karol Kucera            10  16  38.46%
 

gogo

Legend
I was going to say that the top list was loaded with servebots, but then I looked at the second list and I realized it wasn't that simple.

So what do you take from this?
 

metsman

G.O.A.T.
@gogo Though that Atlantic article makes compelling points, Federer can't be too happy with his 20% win percentage, because that is a lot of matches where he didn't capitalize on missed opportunities. For example when he lost the Rome final to Nadal in 5 sets in 2006 where he won more points (179 to 174) and more breaks of serve (4 to 3):
Yeah, I think what hurts much more than the % is the amount of big matches it happened to Federer in. 05 AO/06 Rome/09 AO.
 

George Turner

Hall of Fame
I was going to say that the top list was loaded with servebots, but then I looked at the second list and I realized it wasn't that simple.

So what do you take from this?

This purely indicates how lucky the player is. Nothing else.

it does not indicate how well players play big points and how mentally strong they are etc. because Mental Midget Ivanisevic has more wins than losses this way as does Safin. While Pistol Pete, a great clutch player, surprisingly has more losses than wins. Nor can i see any pattern of a particular type of player benefiting from these matches, it seems quite random. Some servebots and some grinders do well, others not so well.

Federer's amount of losses can be put down to inferior players needing luck to squeeze past him.
 

AnOctorokForDinner

Talk Tennis Guru
This purely indicates how lucky the player is. Nothing else.

it does not indicate how well players play big points and how mentally strong they are etc. because Mental Midget Ivanisevic has more wins than losses this way as does Safin. While Pistol Pete, a great clutch player, surprisingly has more losses than wins. Nor can i see any pattern of a particular type of player benefiting from these matches, it seems quite random. Some servebots and some grinders do well, others not so well.

Losing dozens of close matches is not about luck, sorry.

As for Sampras vs Ivanisevic, it's a curious case because those two were different players in BO5 and BO3. Ivanisevic was no mental midget in ordinary matches, whereas Sampras would often take them relatively lightly. I'm not surprised with Ivanisevic having a better clutch record in BO3. But Pete routinely turned the heat up for the big ones, while Goran would often end up crumbling. It would be useful to separate BO5 vs BO3 clutch record in general, I'm sure you'll see some players who consistently over- or underperform in BO5.
 

AnOctorokForDinner

Talk Tennis Guru
I'm a bit confused. What does the 8 associated to Federer represent?

The number of matches he won while winning fewer points.

This is the MTF thread that has a similar table, but also covers Big 4 in detail: www.mens*tennis*forums.com/2-general-messages/888113-records-when-winning-despite-losing-same-more-points-than-their-rival.html (Note that, in that thread, 50/50 exactly equal matches are also counted, which e.g. adds an extra 3-4 to Federer's record, taking it from 8-32 to 11-36).
 

mike danny

Bionic Poster
8 is number of times he won a match when winning less points than his opponent.

32 is the amount of times he lost a match despite winning more points than his opponent.
The number of matches he won while winning fewer points.

This is the MTF thread that has a similar table, but also covers Big 4 in detail: www.mens*tennis*forums.com/2-general-messages/888113-records-when-winning-despite-losing-same-more-points-than-their-rival.html (Note that, in that thread, 50/50 exactly equal matches are also counted, which e.g. adds an extra 3-4 to Federer's record, taking it from 8-32 to 11-36).

It's shocking how many times Fed has been better than his opponents.
 

George Turner

Hall of Fame
Losing dozens of close matches is not about luck, sorry.

As for Sampras vs Ivanisevic, it's a curious case because those two were different players in BO5 and BO3. Ivanisevic was no mental midget in ordinary matches, whereas Sampras would often take them relatively lightly. I'm not surprised with Ivanisevic having a better clutch record in BO3. But Pete routinely turned the heat up for the big ones, while Goran would often end up crumbling. It would be useful to separate BO5 vs BO3 clutch record in general, I'm sure you'll see some players who consistently over- or underperform in BO5.

Other *mental midgets* with winning records are Monfils, Phillopousis, Nishikori (17-6??) and Berdych. None of those guys you associate with being clutch. That said it would be interesting to see if there's any pattern between BO3 and BO5.
 

AnOctorokForDinner

Talk Tennis Guru
Other *mental midgets* with winning records are Monfils, Phillopousis, Nishikori (17-6??) and Berdych. None of those guys you associate with being clutch. That said it would be interesting to see if there's any pattern between BO3 and BO5.

Nishikori is nothing like a mental midget, get serious. He's got the best deciding set record in the Open era. His glass body causes inconsistent results, and yes, sometimes he crumbles against superior opposition, but so does everyone, and Nishi's no ATG. Nishikori and Raonic aren't really weak mentally, just not good enough overall + injury-prone.

Regarding the other two, again I'd look at BO3 vs BO5, or 'big matches' (Masters/Slam QF and upwards?) vs ordinary matches. It's the mercurial type of player that doesn't fret much in normal circumstances due to their unpredictability, but put them on the big stage, and they will normally crack. So they may have a good clutch record in 90% of the tour, but the most important 10% is too much.
 

NoleFam

Bionic Poster
BP conversion for those 36 matches combined: 110/322. I thought it'd be worse, actually. Still, there are a lot of individual great gems such as: 3/12, 2/10, 3/10, 4/10, 4/14, 4/16, 4/18, 6/19, 4/21, 2/8, 1/6, 1/7, 1/8...

Bingo. I was just getting ready to post this. Instead of showing how lucky his opponents were, it shows how less than stellar Federer was on those break point opportunities.
 
This list does not show clutch players.
It shows players that tend to tank some games/sets in order to play better the next one.
Kyrgios will top this list soon...
 

AnOctorokForDinner

Talk Tennis Guru
Bingo. I was just getting ready to post this. Instead of showing how lucky his opponents were, it shows how less than stellar Federer was on those break point opportunities.

Most certainly, the fact that tennis scoring means some points are more important than others is a massive blow to Federer's career :D If only his performance hadn't dropped on big points so often, those records of his would've already been out of reach.

Funnily enough, it's true that, barring teenage seasons, Federer is the toughest of the Big 3 to shut out on serve in a loss - he has a lot more no-break TB-assisted wins than Nadal & Djokovic, but fewer no-break losses, starting from the first Top 10 season: Fed = 28 since 02, Nads = 37 since 05, Djok = 32 since 07. Yet that has nothing to do with BP and MP conversion issues.
 

George Turner

Hall of Fame
Greatest clutch shot in the history of tennis.

IllInferiorJanenschia-size_restricted.gif
 

George Turner

Hall of Fame
This list does not show clutch players.
It shows players that tend to tank some games/sets in order to play better the next one.
Kyrgios will top this list soon...

That's a better explanation.

Though Kyrgios won't ever be on this list, when he tanks he tanks whole matches.
 

NoleFam

Bionic Poster
Most certainly, the fact that tennis scoring means some points are more important than others is a massive blow to Federer's career :D If only his performance hadn't dropped on big points so often, those records of his would've already been out of reach.

Funnily enough, it's true that, barring teenage seasons, Federer is the toughest of the Big 3 to shut out on serve in a loss - he has a lot more no-break TB-assisted wins than Nadal & Djokovic, but fewer no-break losses, starting from the first Top 10 season: Fed = 28 since 02, Nads = 37 since 05, Djok = 32 since 07. Yet that has nothing to do with BP and MP conversion issues.

Well he's an excellent tiebreak player and probably even more so than Nadal and Djokovic, even though they are no slouches in that category. Also, his serve is excellent and rarely completely off, which is why he has so few bagels against him, so he is tough to break as well. His issue has been on those break points throughout his career which have cost him in some key matches.
 

AnOctorokForDinner

Talk Tennis Guru
Well he's an excellent tiebreak player and probably even more so than Nadal and Djokovic even they are no slouches in that category. Also, his serve is excellent and rarely completely off, which is why he has so few bagels against him, so he is tough to break as well. His issue has been on those break points throughout his career which have cost him in some key matches.

That's true, but the stat I posted was different - it showed that, compared to Nadal and Djokovic, it's tougher to beat Federer without getting broken at all. Although his overall return game is lesser, Fed's amazing at generating some chances against any kind of opponent - probably because of his much-praised variety and sense of timing. But then it goes out the window on BP, time and time again :(
 

NoleFam

Bionic Poster
That's true, but the stat I posted was different - it showed that, compared to Nadal and Djokovic, it's tougher to beat Federer without getting broken at all. Although his overall return game is lesser, Fed's amazing at generating some chances against any kind of opponent - probably because of his much-praised variety and sense of timing. But then it goes out the window on BP, time and time again :(

Yea that's why I was bringing up how great of a tiebreak player he is. He is notorious for creating opportunities in the tiebreak even when he is not returning as well as those two are. The opponent serves great and still ends up losing the set.
 
I'm a bit confused. What does the 8 associated to Federer represent?

8 is number of times he won a match when winning less points than his opponent.

32 is the amount of times he lost a match despite winning more points than his opponent.

Same here. I'm very sleepy right now and the percentage is kind of very confusing.

Aren't these separate matches- the wins and the losses...

Why is it 20%? 8 out of 40? isn't the 32 losses a different category? Why are they lumped together?

8 wins when winning less points and 32 losses when winning more points.
 

AnOctorokForDinner

Talk Tennis Guru
Well he's an excellent tiebreak player and probably even more so than Nadal and Djokovic, even though they are no slouches in that category. Also, his serve is excellent and rarely completely off, which is why he has so few bagels against him, so he is tough to break as well. His issue has been on those break points throughout his career which have cost him in some key matches.

I love how these little TB, BP and MP stats really show where a player stands mentally at the time. Nadal especially has made a perfect example of himself. Look at this:
pre-2007 Nadal: 63% TB won in 05-06, losses while gathering 10+ BPs: 9, matches lost from MP up: 4

2007-2013 Nadal: at least 68% TB won per season except for 07 and 09, both at 57% (was at 15-7 (68%) in 09 after USO, but a bad 1-5 record afterwards), losses while gathering 10+ BPs: 10, matches lost from MP up: 1 (GOATdenko :eek:)

2014+ Nadal: a negative 49% TB record in 2014-16 (35-36 combined), losses while gathering 10+ BPs: 16, matches lost from MP up: 2

Such clear ups and downs.
 

AnOctorokForDinner

Talk Tennis Guru
Bingo. I was just getting ready to post this. Instead of showing how lucky his opponents were, it shows how less than stellar Federer was on those break point opportunities.

By the way, although Djokovic has generally been the least likely to waste a huge heap of BPs than Federer and Nadal, in big matches he's had his share of crappery: 2/10, 3/11, 4/13 (twice), 6/15, 3/17 (twice). The last three came just recently - Wim 16, USO 16 and AO 17. How weird it is to go from holding all four to posting 12/49 BP conversion in the next three Slam losses combined.
 

NoleFam

Bionic Poster
I love how these little TB, BP and MP stats really show where a player stands mentally at the time. Nadal especially has made a perfect example of himself. Look at this:
pre-2007 Nadal: 63% TB won in 05-06, losses while gathering 10+ BPs: 9, matches lost from MP up: 4

2007-2013 Nadal: at least 68% TB won per season except for 07 and 09, both at 57% (was at 15-7 (68%) in 09 after USO, but a bad 1-5 record afterwards), losses while gathering 10+ BPs: 10, matches lost from MP up: 1 (GOATdenko :eek:)

2014+ Nadal: a negative 49% TB record in 2014-16 (35-36 combined), losses while gathering 10+ BPs: 16, matches lost from MP up: 2

Such clear ups and downs.

Very true point and ouch for Nadal in tiebreaks from 2014-2016. I knew it would be lower but still surprised it was that low.

By the way, although Djokovic has generally been the least likely to waste a huge heap of BPs than Federer and Nadal, in big matches he's had his share of crappery: 2/10, 3/11, 4/13 (twice), 6/15, 3/17 (twice). The last three came just recently - Wim 16, USO 16 and AO 17. How weird it is to go from holding all four to posting 12/49 BP conversion in the next three Slam losses combined.

Yea that is terrible for him and honestly, I'm not shocked at Wim 2016 and AO 2017. Even USO 2016 he was wasting a bunch of chances as well. It's true that you can tell how mentally solid a player is at the time by looking at those tiebreak and BP conversion rates.
 

Fedforever

Hall of Fame
I'm curious as to what is considered a good BP conversion rate. Given the advantage lies with the server you would expect a high percentage of BPs to be saved. So what counts as good? 50%?
 

metsman

G.O.A.T.
That is actually the dumbest interpretation of any bit of data I've seen on TTW this year.
High praise considering Meles.

And it's true, unless you have examples to the contrary. It's true for any elite level player but moreso for Federer because of how his game works.
 

Red Rick

Bionic Poster
High praise considering Meles.

And it's true, unless you have examples to the contrary. It's true for any elite level player but moreso for Federer because of how his game works.

This data says nothing about Federer's relative form, or how hard it is to beat him. Obviously, Federer has one of the highest % of points won in an average match, so unless he scores significantly below his average, you're unlikely to beat him, but that's got nothing to do with this data. That's just looking at his winning percentage.


- Federer loses a lot more often when winning more points than vice versa

- If anything, this suggests that if Fed wins about 50% of points in a match, he's more likely to lose than to win, and that he wins matches by outscoring them convincingly. Ergo, Federer isn't that clutch in matches that are super close.

- That said, I think it's also misrepresentation of data. Federer wins over 80% of his matches, and he likely wins more points than his opponent in over 80% of his matches as well, so he basically gets more chances to lose when ahead compared to getting chances to win when he loses more points, though looking at data of other players, this doesn't really seem to play a role.

- I think you could describe likelihood of winning a match as a function of % of points won in a match, pool all players, and then you should really be able to compare players to the rest of the population.


Given a few of the players who have a very high winning ratio, namely Nishikori, Sock and especially Fognini, I think this says a lot more about players tanking sets than anything else. Ergo, Federer don't tank no sets, or players tank sets vs Federer relatively more
 
Top