Requests for proper speed tests that would attempt to measure the court speeds of different surfaces relative to one another, are sensible and reasonable requests. A brief reflection on what these tests would involve to be properly conducted, makes me be pessimistic that they will ever be conducted.
First you define what you need to measure. Let’s say it is the horizontal speed of the ball, post-bounce, off various different shots on different surfaces.
Some essential things in no particular order:
- devise a good enough variety of shots to be tested. Two or three different kinds of topspin shots at different rpms, different speeds, and different angles when leaving the ball machine. Same thing with flat shots, underspins, and serves.
- ensure that the exact same machinery is used in all tests and calibrated prior to each test so that the parameters (speed of shot of the machine, angle, rpm) are indeed the same, as well as the measuring techniques for the horizontal speed post bounce.
- ensure that atmospheric conditions that vary from day to day (temperature, humidity, wind and air pressure) and that affect the speed and trajectory of the ball, are the same on all tests. This may well be the most difficult condition to attain, next to impossible, and the only way to compensate for its impossibility would be to increase the sample size under various conditions hoping they would cancel each other out.
Though such comparative tests would not be impossible in principle, they would be expensive to carry out, and they are very unlikely. On the other hand, if they are not conducted properly, and only a few shots are measured without regard to all these factors, the potential for large skewing would be too large for them to have much validity.
Now, in the absence of such tests, there is simply no objective measure to rank court speed directly. Anecdotal evidence that so and so said he felt the courts were faster or slower when he played so and so on such and such day are really not what you call “hard data”. They are in fact pretty meaningless as a measure of anything.
So in the absence of any reliable direct speed tests, it seems to me the next best measure is some kind of proxy measurement that can rely on hard objective data and be free of subjective impressions. One can think of several such proxy measurements, among them length of points, percentage of points won by the receiver, and percentage of games won by the receiver.
The easiest of these to compute based on existing data is the latter. Demonstrating that there is a general relation between percentage of breaks and court speed is not difficult. Which surfaces are faster and which are slower in general, is already known, even though without precision. This we call generally perceived speed. You would need to be slightly deranged to argue this carries no weight and that MC might just be faster than Lyon, for all we know, etc. Since the general relation between perceived court speed and breaking frequency has long been known intuitively and can easily be confirmed through comparisons of breaking percentages on different perceived court speeds (again, by perceived court speeds I mean court speeds whose ranking relative to one another is generally accepted as known; for example, the fact that clay is generally slower than grass, etc.) then there is no question but that breaking percentage can act as a fairly reasonable and reliable proxy measurement for this variable. In fact, it is probably more reliable than many others that are routinely used on other fields:
http://en.wikipedia.org/wiki/Proxy_(climate)
http://en.wikipedia.org/wiki/Proxy_(statistics)
Finally, I will mention that those who continue arguing against the lack of objectivity or "science” in using breaking percentage as a proxy indicator of court speed, have themselves nothing whatsoever to offer in the way of objectivity and "science", except purely anecdotal trivialities based on the “feeling” of such or such player, or on their own impressions, or on the fact that such and such player made it to such and such round. Not particularly impressive.
Even more absurd is the supposition that these analysis are prompted by a need to defend or attack particular players.
When these discussions about Wimbledon "turning to clay" began a few years ago, I quickly realized there were no reliable available direct measurements to confirm this. I also realized there was an enormous amount of confusion about what is meant by “court speed”. It was then I began to think of possible proxy indicators that could shed some light away from all that muddy chatter. I first thought that the percentage of points won by receivers vs total points played, over entire tournaments year after year, would be a pretty good indicator. But since this data is not readily available, obtaining in the necessary amounts to get large sample sizes would be extremely lengthy. Next best thing would be to focus on games won by receivers vs total games. When NF presented his first set of results on this, I was in a sense not surprised to see the correlation with perceived court speeds easily confirmed. But on the other hand I was surprised at how clear and smooth the correlation was when using these very large samples. This completely convinced me that, in the absence of those direct tests, breaking percentage is by far the best available proxy indicator of this variable.
Those who deride it as outlandish and “unscientific” should have better things to offer than the reproduction of anecdotal comments about the impressions some players or people have.
First you define what you need to measure. Let’s say it is the horizontal speed of the ball, post-bounce, off various different shots on different surfaces.
Some essential things in no particular order:
- devise a good enough variety of shots to be tested. Two or three different kinds of topspin shots at different rpms, different speeds, and different angles when leaving the ball machine. Same thing with flat shots, underspins, and serves.
- ensure that the exact same machinery is used in all tests and calibrated prior to each test so that the parameters (speed of shot of the machine, angle, rpm) are indeed the same, as well as the measuring techniques for the horizontal speed post bounce.
- ensure that atmospheric conditions that vary from day to day (temperature, humidity, wind and air pressure) and that affect the speed and trajectory of the ball, are the same on all tests. This may well be the most difficult condition to attain, next to impossible, and the only way to compensate for its impossibility would be to increase the sample size under various conditions hoping they would cancel each other out.
Though such comparative tests would not be impossible in principle, they would be expensive to carry out, and they are very unlikely. On the other hand, if they are not conducted properly, and only a few shots are measured without regard to all these factors, the potential for large skewing would be too large for them to have much validity.
Now, in the absence of such tests, there is simply no objective measure to rank court speed directly. Anecdotal evidence that so and so said he felt the courts were faster or slower when he played so and so on such and such day are really not what you call “hard data”. They are in fact pretty meaningless as a measure of anything.
So in the absence of any reliable direct speed tests, it seems to me the next best measure is some kind of proxy measurement that can rely on hard objective data and be free of subjective impressions. One can think of several such proxy measurements, among them length of points, percentage of points won by the receiver, and percentage of games won by the receiver.
The easiest of these to compute based on existing data is the latter. Demonstrating that there is a general relation between percentage of breaks and court speed is not difficult. Which surfaces are faster and which are slower in general, is already known, even though without precision. This we call generally perceived speed. You would need to be slightly deranged to argue this carries no weight and that MC might just be faster than Lyon, for all we know, etc. Since the general relation between perceived court speed and breaking frequency has long been known intuitively and can easily be confirmed through comparisons of breaking percentages on different perceived court speeds (again, by perceived court speeds I mean court speeds whose ranking relative to one another is generally accepted as known; for example, the fact that clay is generally slower than grass, etc.) then there is no question but that breaking percentage can act as a fairly reasonable and reliable proxy measurement for this variable. In fact, it is probably more reliable than many others that are routinely used on other fields:
http://en.wikipedia.org/wiki/Proxy_(climate)
http://en.wikipedia.org/wiki/Proxy_(statistics)
Finally, I will mention that those who continue arguing against the lack of objectivity or "science” in using breaking percentage as a proxy indicator of court speed, have themselves nothing whatsoever to offer in the way of objectivity and "science", except purely anecdotal trivialities based on the “feeling” of such or such player, or on their own impressions, or on the fact that such and such player made it to such and such round. Not particularly impressive.
Even more absurd is the supposition that these analysis are prompted by a need to defend or attack particular players.
When these discussions about Wimbledon "turning to clay" began a few years ago, I quickly realized there were no reliable available direct measurements to confirm this. I also realized there was an enormous amount of confusion about what is meant by “court speed”. It was then I began to think of possible proxy indicators that could shed some light away from all that muddy chatter. I first thought that the percentage of points won by receivers vs total points played, over entire tournaments year after year, would be a pretty good indicator. But since this data is not readily available, obtaining in the necessary amounts to get large sample sizes would be extremely lengthy. Next best thing would be to focus on games won by receivers vs total games. When NF presented his first set of results on this, I was in a sense not surprised to see the correlation with perceived court speeds easily confirmed. But on the other hand I was surprised at how clear and smooth the correlation was when using these very large samples. This completely convinced me that, in the absence of those direct tests, breaking percentage is by far the best available proxy indicator of this variable.
Those who deride it as outlandish and “unscientific” should have better things to offer than the reproduction of anecdotal comments about the impressions some players or people have.
