Mens Tennis Forums banner

1 - 9 of 9 Posts

·
Registered
Joined
·
28 Posts
Discussion Starter #1
Hi Guys,

I was toying with the idea of correlating some stats for the top 100 players.
In particular I always found unsatisfactory the "standard" stats that are used to
evaluate the strength of a player serve and wanted to find out things like the
how many times (%) a given player holds serve against top 10/20/50 players, because
I think it would be a better indicator of the serve strength.

Well, does anybody know where to find the actual data? Like in a text format or whatever?
ATP and CORE don't quite give data away so easily. Or maybe I didn't find it.
 

·
Registered
Joined
·
14,409 Posts
Only of 2007 or in general?

It's actually pretty difficult to find stats, which are more complexe than those on atptennis.com. I really find them not that bad (compared to wtatour.com for example) and think you won't find better ones anywhere else, maybe stevegtennis.

If you're only looking for 2007 stats, you can for example take the match details in the player profiles as a basis.
 

·
Registered
Joined
·
100 Posts
spot checking atptennis, they seem to have individual match stats back through 1991. of course using these is a very time intensive, brute force and awkwardness type of task.

i like your idea of compiling stats based on ranking of opponent. given the unavailability of a database that can be queried doing it for top 100 seems unrealistic though.

one of the stats that interests me is a player's record against top 10 opponents. anyone have a compilation of this or know where to find it? this caught my interest when some digging turned up the realization that davydenko and robredo fare poorly and some other top 10 players recently have only beaten these two, etc.


what/where is CORE, btw? my conversion from casual fan to obsessed fanatic is fairly recent so i'm still unaware of some of the resources. thanks.
 

·
Registered
Joined
·
28 Posts
Discussion Starter #5
I used the data from the ATP web site. Since the information is not available as a single data file, I couldn't do anything extensive and I just compared the serve stats of Federer, Roddick and Karlovic.

The outcome is interesting. Here it is.

A) If we look at overall "Service Games Won %" we find:
1. Karlovic 94%
2. Roddick 91%
3. Federer 89%

This suggests the usual lore according to which, for what concerns the serve, Kar > Rod > Fed

B) If we look at the "Service Games Won Against Top 20 Players %" we find:
1. Karlovic 94.3% ( 14 matches played, average ranking of opponent= 8.1 )
2. Federer 87,2% ( 32 matches played, average ranking of opponent= 9.2 )
3. Roddick 83,1% ( 20 matches played, average ranking of opponent= 8.9 )

Ah ah! This would suggest that if we take grossly into account the opponent strength (i.e.
we look only at top 20 and discard everybody else), we obtain Kar > Fed > Rod!

But it is not over.

C) If we are more subtle in considering the opponent strength and construct the following stat:
"Service Games Won Against Top 20 Players %" x "Average Ranking of Opponent", we find:
1. Federer 8.0
2. Karlovic 7.7
3. Roddick 7.4

This would suggest that if we take into account the quality of the opponent, as incredible as
it is, Federer has a "higher" probability of winning a service game than Karlovic.

CAVEATS:

1. This doesn't mean that Federer serve is better than Karlovic one, but that his overall game-play in his service games is more effective than Karlovic one. So If you replace Federer with Karlovic for the service games you would have a less winning player.

2. Better would be to construct the returning stats for the top 20 players and using those stats to weight the opponent strength (unfortunately in some model-dependent way) when studying the Serve percentages.

I leave the latter for future work.

Enjoy the new (minor and meaningless) insight...

Snoo
 

·
Registered
Joined
·
1,639 Posts
C) If we are more subtle in considering the opponent strength and construct the following stat:
"Service Games Won Against Top 20 Players %" x "Average Ranking of Opponent", we find:
1. Federer 8.0
2. Karlovic 7.7
3. Roddick 7.4
That doesn't make sense: Your formula gives more points to a player whose opponents are weaker?
...As in previous examples Karlovic had highest win% and highest ranked opponents in top 20. So surely he should be on top at "C" too.
 

·
Registered
Joined
·
1,451 Posts
I used the data from the ATP web site. Since the information is not available as a single data file, I couldn't do anything extensive and I just compared the serve stats of Federer, Roddick and Karlovic.

The outcome is interesting. Here it is.

A) If we look at overall "Service Games Won %" we find:
1. Karlovic 94%
2. Roddick 91%
3. Federer 89%

This suggests the usual lore according to which, for what concerns the serve, Kar > Rod > Fed

B) If we look at the "Service Games Won Against Top 20 Players %" we find:
1. Karlovic 94.3% ( 14 matches played, average ranking of opponent= 8.1 )
2. Federer 87,2% ( 32 matches played, average ranking of opponent= 9.2 )
3. Roddick 83,1% ( 20 matches played, average ranking of opponent= 8.9 )

Ah ah! This would suggest that if we take grossly into account the opponent strength (i.e.
we look only at top 20 and discard everybody else), we obtain Kar > Fed > Rod!

But it is not over.

C) If we are more subtle in considering the opponent strength and construct the following stat:
"Service Games Won Against Top 20 Players %" x "Average Ranking of Opponent", we find:
1. Federer 8.0
2. Karlovic 7.7
3. Roddick 7.4

This would suggest that if we take into account the quality of the opponent, as incredible as
it is, Federer has a "higher" probability of winning a service game than Karlovic.

CAVEATS:

1. This doesn't mean that Federer serve is better than Karlovic one, but that his overall game-play in his service games is more effective than Karlovic one. So If you replace Federer with Karlovic for the service games you would have a less winning player.

2. Better would be to construct the returning stats for the top 20 players and using those stats to weight the opponent strength (unfortunately in some model-dependent way) when studying the Serve percentages.

I leave the latter for future work.

Enjoy the new (minor and meaningless) insight...

Snoo
Good effort, but faulty reasoning near the end. Multiplying the percentage by the ranking actually gives the opposite of what you want, as a higher ranking number indicates a lower position on the ranking (sounds weird, but what I mean is that the world number 1 is ranked HIGHER than the world number 20, while 20 is a larger number than one - that's what I mean by "higher ranking number"). What you could do is devide, but that could also result in very strange results (for instance, a 100 % service points won record would result in a 1 percent/ranking place figure if the average ranking of the opponents was 100, whereas an 80 % service points won record would result in a 4 percent/ranking place if the average ranking of the opponents was 20, indicating that the latter is the better server when this clearly is not the case - however, if the average rankings of the opponents are fairly close, this could work to provide a rough guideline). Bottom line is, you need a more advanced model that takes into account how much more a player loses as the ranking of the opponent gets higher - a simple multiplicative relationship will not capture this.
 

·
Registered
Joined
·
9,760 Posts
Service games won % is a noisy metric for serve ability because it also carries out information about ability to win points after the serve. For example, you see that Nadal appears as #5 in the category and I doubt that anybody would make a case for him being the #5 server on tour.
 

·
Registered
Joined
·
1,451 Posts
Service games won % is a noisy metric for serve ability because it also carries out information about ability to win points after the serve. For example, you see that Nadal appears as #5 in the category and I doubt that anybody would make a case for him being the #5 server on tour.
Absolutely correct. A good rallier who wins most games will win lots of service games too. We have to isolate the impact of the serve as such.
 
1 - 9 of 9 Posts
Top