Shot-Stats, Babolat Play, Zepp, etc - let's discuss!

Discussion in 'Other Equipment' started by v-verb, Mar 19, 2014.

  1. v-verb

    v-verb Hall of Fame

    Joined:
    Nov 9, 2012
    Messages:
    2,077
    Location:
    Toronto
    I'm interested in all these feedback devices. Just want to have a place to discuss features and benefits of each.

    Anyone know when an APD Play is coming out? I may jump on that. Looking forward to checking out Shot-Stats when it's ready as well.

    I'm sure the other big racquet companies will respond - any news on this?

    Cheers
     
    Last edited: Mar 19, 2014
    #1
  2. corners

    corners Legend

    Joined:
    Jul 31, 2008
    Messages:
    5,441
    First of all, I think it would be very useful to discussions of these things if we had a basic understanding of how they worked. What exactly are they measuring? Vibration? How are the vibrations processed and interpreted?

    It would be great to collect a list of links to any information about what's "under the hood" of these devices as published by the manufacturers. I suspect they are all being kind of secretive, but maybe not, I haven't looked into it. My feeling is that until we know how these things work we can't really discuss them intelligently, nor compare the various products against each other.
     
    #2
  3. ShotStatsOfficial

    ShotStatsOfficial New User

    Joined:
    Aug 9, 2013
    Messages:
    64
    Hey there,m corners!

    Lavie here, one of the founders from Shot Stats and here to answer your tennis questions. Sergey, our other founder, specializing on the engineering side, will also be hear to answer any engineering questions.

    Looks like this first one is a good one for Sergey. He'll be able to shed a bit a light for you.

    Thanks!
     
    #3
  4. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    Thanks for the thread v-verb!

    Corners, Sergey here from Shot Stats. I'm the lead engineer/general tech guy/mountain man. :)

    Boring tech stuff ahead!

    I'd be happy to give you some info about what exactly is under the hood but at this point, being in active development, it would be slightly premature.

    To answer your question about what exactly is being measured:

    1) Bulk motion - the overall movement of the racket such as during a swing.
    2) Vibration.

    Pretty much all of the information comes from those two things. Physically, both of them come down to measuring one thing: acceleration, but on different time scales. It's difficult to do both at the same (with one sensor) because the sample rates necessary are very different. This is why both Babolat and Zepp use multiple sensors.

    As an example, something like shot count seems pretty simple at first - just count whenever you detect an impact. Unfortunately differentiating between an actual ball impact and someone bouncing a racket on their knee between points is pretty hard. To do it well it'd be best to use both motion and vibration data. The motion data helps you see if there was something resembling a proper swing before an after the impact and - if so - will count the shot. This is also why volleys are so difficult to detect - there is not a lot of racket motion you can use to firmly categorize the shot.

    It's all easier said than done. There are PhD theses and research papers written about this very thing in tennis! We have our theories of how Babolat and Zepp do it based on using both products, but I don't think it would be appropriate for me to speculate on a public forum. Especially one that's been so accommodating to all of us. Suffice to say that what they're doing is truly awesome and we're excited to be a part of it.

    When it comes to the kinds of sensors (accelerometers/gyroscopes) used in these products, the truth of the matter is that they are all more or less the same. There are not all that many options for sensors. They all do the same thing, and you can bet that any company like Babolat, Zepp, or even small fries like us are going to use the best sensors available to them. The hardware is relatively easy, and for 95% of all fitness or sports trackers (Nike Fuel, Fitbit, Withings Pulse, etc...) is pretty much the same on the inside. The real magic comes from the software - the analysis you do with that raw data. This is why most companies doing this kind of thing, for tennis or otherwise, are not too eager to discuss the fundamental workings. Those are the secret sauce you mentioned. Having said that there are definitely advantages to getting creative with hardware as we hope to do. ;)

    I apologize for the wall of text, I'm a wordy person by nature. :) I hope I could shed a bit of light on your questions. If not I'm happy to elaborate on any particular point. Having said that, I don't want to monopolize this thread so I'll try and stay out of the way unless there are specific questions for me.

    Thanks and all the best!
     
    #4
  5. magmasilk

    magmasilk New User

    Joined:
    Aug 29, 2006
    Messages:
    84
    Location:
    Wash, DC
    Sergey,

    Thanks for the insight. I'm currently using zepp and always interested in understanding what is it and other types thing really measuring and then trying to describe.

    Any thoughts on the power numbers on these types of systems you can share? Zepp seems to be relative to "Pro" like babolat. Also there was some speculation on the board but the incoming speed doesn't really seem to matter. I haven't done extensive "testing" but hand feed and none hand feed don't seem to change the numbers notably.

    Also here are my thoughts on volley. I'm working on the assumption that shot under spin/ flat slices could be confused as a volley. I say make that coding be user defined as in know for me, I volley more than slice.

    Thanks for the info.
     
    #5
  6. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    It's really hard to say. There are a number of ways they might get a "power" number. Could be the magnitude of the shock when hitting a ball, could be the speed of the racket head on contact, could be something else entirely, or a combination of all of them! The dynamics of a tennis racket are quite complex. There are so many things going on and so many ways to extract useful information out of them. Nailing down what - exactly - they are measuring and how they arrive at a power number is difficult to do without some sort of mechanical test rig that could guarantee shot-to-shot consistency. What you said - that the incoming ball speed doesn't seem to make a difference - seems to indicate that the motion of the racket itself and not the impact is used to calculate power.

    If you've seen Zepp's golf app, you know it can show the swing of a golfer. In many ways it is relatively "easy" to do this for golf. The player is stationary and the swing has a few well defined phases. You bring the club back, stop, swing, hit the ball, follow through, stop. It's a very "clean" motion - there's nothing going on besides the swing. In tennis, on the other hand, you're in constant motion. In the midst of all this fast movement it's much harder to pick out relatively small events like volleys.

    So, keeping stroke information clean without "contaminating" it on account of a player's movement is quite hard. Imagine a line traced through space by a golf club during a single swing. Now imagine a line drawn by the head of a racket throughout a point! It would look like chaos in comparison. This is probably why you see a lot of this kind of 3D visualization in golf, but not so much in tennis. Not yet, anyhow. It's a much more difficult problem to solve.
     
    #6
  7. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    In fact, Zepp support has told me that on the horizon is an update which will provide a 3D tennis serve swing analysis.
     
    #7
  8. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    In fact, Zepp support has told me that on the horizon is an update which will provide a 3D tennis serve swing analysis, this being more similar to a static golf swing.

    As far as power is concerned, since F=Ma, couldn't it be derived from the "delta a" post impact?
     
    Last edited: Mar 19, 2014
    #8
  9. KineticChain

    KineticChain Professional

    Joined:
    Feb 26, 2012
    Messages:
    1,158
    Just a fun fact; change in acceleration is called 'jerk'. Just curious, what would F=ma have to do with just detecting acceleration? Maybe a quick change in the accelerometer's reading during the impact can be scaled with a constant for a 'power number', but I don't think the mass can be considered a constant with swing dynamics. The arm has hinge points at elbow and wrist, which would affect the mass that the ball see's during impact. It's like swinging a 10 kg steel rod at 30 mph vs swinging a 10 kg steel chain at 30 mph. Since the chain is not as rigid as the solid steel rod, a lot of impact energy will be dissipated through the hinging links. The same concept can be applied to swinging at the ball with your arm. A more fundamentally sound technique that does not allow elbow and wrist hinging during impact will put more momentum into the shot. I think contrary to what you always hear on the forums, the racket head speed is not the only factor when considering the power of a shot.

    TL;DR: I think the mass the ball "sees" depends on how you swing at the ball
     
    #9
  10. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    F=ma is not for detecting acceleration. You have the acceleration measured just before impact, and then the lesser acceleration after impact (slight rebound). The loss of acceleration times the swingweight should deliver the power of the shot. This could be problematic though because the force of impact will also be affected by the speed of the incoming ball. Golf is a lot easier since the golf ball is stationary. I wonder how the speed of the ball is taken into account...
     
    #10
  11. KineticChain

    KineticChain Professional

    Joined:
    Feb 26, 2012
    Messages:
    1,158
    Yes I know F=ma is not for detecting acceleration, I thought you were implying that with your original post. Swingweight is an attribute of a racket, right? What is its physical meaning?
     
    #11
  12. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    Pick up a racket, and it will feel heavier or lighter depending on where we hold it. Swingweight is an objects resistance to change in its rotation. Mass in circular motion is what swingweight is about.

    Depending on the distribution of its mass, a racket's swingpath will be the sum of each bit of mass times the distance squared to the pivot point of rotation. So you can see how the weight distribution on the racket affects the swingweight: adding lead towards the head of the racket increases the swingweight much more than adding lead near the handle.

    A racket with a large swingweight will be difficult to swing and vice versa.
     
    #12
  13. Dimcorner

    Dimcorner Professional

    Joined:
    Oct 12, 2012
    Messages:
    814
    I would think a sensor would register an impact on the string bed as an event and you would save the data from about under 1 second prior to about 1/4 second after to get racquet head orientation and trajectory data. You can probably discard shots when the racquet is parallel to the ground with no horizontal acceleration (just bouncing a ball before serve) or if the 2+shots come within 1 second of each other. This should cover like +95% of your shots no?

    Wonder if you can use a ball bounce to tell the sensor that a serve is the next shot to be hit... Sure would be easy that way to let it know it's a serve instead of overhead.

    I really hope the shot-stats sensor comes in at about a dampener or 2 in weight and has enough juice to last a whole match (3 hours give or take).
     
    Last edited: Mar 20, 2014
    #13
  14. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    Dimcorner, you hit the nail on the head. :) Those are some of the ways you can reject false "shots," and part of what we are incorporating for that purpose. Finding the trajectory of the racket is fairly involved but is definitely the best source of information to figure out what is and is not a valid shot. Time between shots and racket orientation are also integral pieces of info.

    As far as weight is concerned, our "hard" goal is 20 grams and our desired goal is closer to around 15. Of course, lower is better. There is a balance to be found because there are a lot of things conspiring to make the thing heavier. The battery is a big one:

    1) Enabling the device to be used standalone requires running most of the processor-intensive calculation onboard.
    2) More powerful processors consume more power.
    3) Higher power consumption means either a bigger battery (thus heavier) or a shorter life.

    The preference for lighter weight vs. longer life might change for different players, but less weight is the primary goal. Around 4 hours of battery life should be a good compromise.

    It's worth mentioning that the Zepp is really a marvel of miniaturization at only 8 grams. We were pretty impressed with the hardware when we first got a chance to play with it.
     
    #14
  15. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    A TW review still not available

    A TW review of Zepp still not available.
    I am NOT talking about testing by members of this Forum
     
    #15
  16. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Racket Head Speed

    Greetings,

    Btw:
    A proof of concept video is suggesting that shot-stats video will measure
    racket head speed
    It is not obvious what a Zepp sensor does
    The base system to compare with should be a system based on reflective markers
    Some references to the latter are provided in the paper by Reid,Campbell and Elliot
    Journal of Applied Biomechanics,2012,28,93-98
    The full paper available from www.ResearchGate
    Julian
    PS It is believed that coaches do observe racket head speed
     
    Last edited: Mar 24, 2014
    #16
  17. TW Staff

    TW Staff Administrator

    Joined:
    Jan 21, 2004
    Messages:
    24,539
    Location:
    Tennis-Warehouse.com
    julian,

    We are waiting for a software update from Zepp before we give our review. We don't have an estimate time for the review yet. Thank you for your patience.

    Brittany, TW
     
    #17
  18. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland

    Reflective markers to capture motion certainly give you more points to measure with a high speed camera, so that you can measure all parts of the body's kinetic chain plus the racket, but it is a very complex measuring system.

    So this is much more information than tennis sensors give you, since the role of hip rotation, pronation, upper arm rotation, etc. can be detected. The setup and equipment needed is much higher cost and beyond the budget of tennis players.

    The sensor is a low cost approach which just delivers motion measurement from one specific point, usually the butt of the racket.

    So, in this respect, unless you wear a lot of sensors, you will not get nearly as much information as with reflective markers.

    The sensors, however, can measure the impact power directly. Not sure how the reflective sensors would do that. Grip strength plays a role in power transfer to the ball, for example.
     
    #18
  19. v-verb

    v-verb Hall of Fame

    Joined:
    Nov 9, 2012
    Messages:
    2,077
    Location:
    Toronto
    Learning a lot - thanks guys!!
     
    #19
  20. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Testing sensors

    The question I was trying to introduce is how do you trust values produced by
    sensors?
    How do we know that a value of topspin is calculated correctly?
     
    #20
  21. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    Hard to say. Not easy, since it also depends on the spin of the incoming ball, string tension, swing path, racket face angle... probably just an approximation.
     
    #21
  22. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    The unfortunate truth is that it's impossible to directly measure the spin of the ball from on-racket sensor data. The best way to generally do this is to use reference points and a high speed camera, but of course this is a serious investment of both time and set up.

    Any information gleaned from the racket itself will be an estimate. The real question is how good of an estimate can be made. That is, of course, what we intend to find out. There is a dizzying array of factors to consider, including string tension, string material, string shape, condition of the ball, etc... The good news is that there is a lot of in-depth research in this field that can be used to make the best estimates possible. Our goal is not to measure the actual spin of the ball coming off the racket, as that also requires knowledge about the spin of the incoming ball, but to measure the estimated change in spin applied to a ball.

    When it comes to racket head speed, this is something that can be measured more or less directly by on-board sensors. The problem then becomes filtering and processing the data to output the correct information.

    Julian, to your question about whether the data can be trusted:

    The short answer is that no sensor data from any system can be fully trusted without comparing it to a known reference. In this case, with racket head speed and ball spin, that reference will be measurement taken from high speed video. If a particular system has shown itself to be accurate in a variety of situations when compared with a known reference, it's probably safe to trust it just a little bit. :)
     
    #22
  23. v-verb

    v-verb Hall of Fame

    Joined:
    Nov 9, 2012
    Messages:
    2,077
    Location:
    Toronto
    Sergey,

    Without causing a problem, is Babolat's system similar? If you can't answer that I understand

    Cheers
     
    #23
  24. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    The input of Lavie is missing

    Sergey,
    A basic scenario is like that:
    I do have a student with a sensor.
    I get a power number for ,say.forehand averaging 100 shots
    Next I change mechanics of a stroke.
    I get new 100 power numbers.
    I calculate an average and compare a new average with an old average.
    Say the difference is 10%
    How can I "trust" that my student is improving?
    Is it possible that I was lead astray by the NOISY sensor?
    Please see as well the title of the post
    Please note that I eliminated all equipment variables-
    say that I am interested in skills of my student only
     
    Last edited: Mar 26, 2014
    #24
  25. v-verb

    v-verb Hall of Fame

    Joined:
    Nov 9, 2012
    Messages:
    2,077
    Location:
    Toronto
    #25
  26. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Important angles for serve

    Please see videos in www.somaxsports.com
    Click the icon "motion analysis"
     
    Last edited: Mar 28, 2014
    #26
  27. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    #27
  28. v-verb

    v-verb Hall of Fame

    Joined:
    Nov 9, 2012
    Messages:
    2,077
    Location:
    Toronto
    Thanks for the link. I read it and am a bit dismayed that Babolat Play is having issues. I'm sure they'll fix it hopefully soon.
     
    #28
  29. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    #29
  30. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    Interesting article gzh. I swear I read that same article a couple weeks ago. Wish my memory was better, can't recall where it was!

    Interesting video from Somax as well. This is one area where sensors would really benefit analysis techniques, as it can be hard to get very precise measurements of things such as hip rotation without a comprehensive camera setup and ideally multiple perspectives.

    Julian, to your point about noisy sensors:

    It is true that sensors (specifically, accelerometers) are inherently noisy to varying degrees. Gyroscopes and magnetometers are much less noisy and in fact any reasonably modern motion tracking system (to include Zepp, Babolat, and things like quadcopters and nearly every smartphone) uses a combination of data from different sensors to achieve better accuracy and consistency than you could get from any individual sensor. "Sensor fusion" is the term of the day here, and it's a common technique to get the most out of any integrated sensor package.

    Dealing with this noise requires some fairly sophisticated but common signal processing methods. The effect a "noisy" sensor would have on the processed output data would be, at least in this application, far less than 10%. That's for a single data point, or stroke. Over 100 strokes, random sensor noise would be averaged out to be negligible. So it's extremely unlikely that a 10% difference between two sets of 100 strokes would be caused by sensor noise. If there was indeed a 10% error over that many strokes, the culprit would be somewhere in the data processing chain and could be found and corrected.

    In short, the raw sensor noise isn't the main concern there. To give you an idea, the error in the raw data due to noise for one of our sensors is in the neighborhood of 0.025%.
     
    #30
  31. Shroud

    Shroud Hall of Fame

    Joined:
    Apr 17, 2013
    Messages:
    3,836
    Location:
    Bay Area
    #31
  32. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Swingpath of a double handed backhand

    Sergey,
    I have another coaching question about sensors.
    Please note the word "coaching"
    Let us drop for a moment a possible improvement via power numbers.
    Say a swingpath of a double handed backhand is described in
    https://www.youtube.com/watch?v=lPBM-hE1Dhw
    Question:
    A trajectory of a tip is described by a yellow line.
    How much of a yellow "line" can be "described" by sensors?
    A question is NOT a science fiction because Zepp is getting closer to provide some 3D graphical description
    of serve.
    So one may hope that a graphic description of a double handed backhand will be available
    at some moment (say from Zepp)
    Regards,
    Julian W.Mielniczuk
    PS One possible parameter is steepness of a forward part of a swing.
    I understand that I can get this from a video as well but it is not the point here.
    In the case of serve we have 4 potential parameters-please see the thread
    "Speed and ave depth of SW ..." in tennis tip/instruction post number 30
     
    Last edited: Mar 30, 2014
    #32
  33. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Racket head speed ?

    There was a conjecture that power numbers produced by Zepp
    are really racket head speeds.
    Somehow nobody jumped on this wagon.
     
    #33
  34. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    Julian,

    That 3D visualization you mentioned is one of our objectives. It's not science fiction at all, it's entirely possible to get that data from the sensors. The hard part isn't so much getting the data but interpreting it correctly. As I mentioned earlier, it's easier to do for golf because golfers are stationary during a swing and you can be confident that all of the motion is caused by the swing. Tennis is different - tennis players are in constant motion during a match and this makes it more difficult to determine which part of the motion is the stroke and which part is actually player movement about the court.

    However it is certainly do-able and we are confident that this kind of functionality will exist very soon, certainly within a year or so. Whether it's done first by Babolat, Zepp, Shot Stats, or someone else remains to be seen. ;)

    As for Zepp's power number, I can't speak to what exactly it is based on. If it is indeed racket head speed that should be straightforward to measure and compare.
     
    #34
  35. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Easiest cases

    Sergey,
    it is obvious that there are easiest cases to interpret-
    one of them is a platform based 3D serve
    The word platform meant to be an opposite of a term "pinpoint"
    A decomposition of a racket head speed into three separate component is of some value in the case of
    both serve and forehand.
    It is believed that a vertical component of RHS is significant for a topspin serve.
    A shoulder rotation (measured in degrees) is a reasonable parameter to provide as well
    Julian W.Mielniczuk
     
    Last edited: Mar 31, 2014
    #35
  36. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Zepp related links

    https://play.google.com/store/apps/details?id=com.zepp.ztennis
    It contains a review if one clicks ">"
    It does work on an ipad,does NOT work on my laptop (double click should work,I think)
    ----->the review (NOT mine) below

    Tung, Aaron Nguyen Son
    First impression not very impressed. Just did a few swings in my room with this new sensor. Seems like the swing needs to reach certain level power/ball impact in order to register. And some how it registered some of my forehand as smash. We test it much more in the upcoming matches. Will revise the review again. Also, there is a small bug: In Power -> Serve, the title is 'Sessions', which is wrong. Update: Tested for a few matches. Have some findings and questions: 1) App in android always hang and crash when syncing. Open again and it seems it still synced successfully though. 2) Serve does not have breakdown topspin/flat. 3) Volley? 4) What is the unit for power?
     
    Last edited: Mar 31, 2014
    #36
  37. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    I am NOT only one confused

    Please see the sentence
    --->
    4) What is the unit for power?
    ---->

    in the review "posted" in post #38
     
    #37
  38. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    There are of course some teething problems cropping up but that is not entirely unexpected as this is a new development in tennis. What goes on behind the scenes is very complicated and we are seeing some of those growing pains in reviews such as the one you mentioned.

    As to the particular comment by the reviewer about the units of "Power," it's really not a question I can answer on behalf of Zepp or Babolat. There's undeniably confusion about what these units are, what they mean, and what they're measuring. Unfortunately, without inside knowledge, it'd purely be speculation on my part.

    In our case, the surest way to build confidence in the data is to compare it to a known reference. This can include things like high speed camera footage, or test rigs that can move the racket in a precise and known path. The data output can then be compared to this reference to determine how accurate it is in a particular situation. This is a type of testing we will be relying heavily on, and we will be sharing the results of that testing as it comes.
     
    #38
  39. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    I think we should put things in perspective, and not have over-inflated expectations to what sensor-based technology can deliver. I am sure it has its limitations, and can not deliver that the information reflector-camera based technology does.

    This is not to say some useful information is upcoming:

    Zepp told me that next in the pipeline is location of ball on racket impact stats, as well as 3D serve swingpath (easier since the serve does not move much, and they have experience with golf).
     
    #39
  40. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    References

    In the case of Joe Shmoe the tennis player references are almost impossible to obtain.
    Someone was trying to be a smart *** and said:
    "Compare Zepp results for serve with numbers from a radar".
    The only problem is that we know that a radar does measure a speed of a ball.
    But we do not know what Zepp does.
    So we do not know whether we compare apples with apples.
    I think it stinks completely.
    We are NOT at the level comparing A to B.
    We are at the mercy of a hardware outfit.
    I expect to be banned from this forum sooner or later.
    It is OK.
    We are talking basically about automating a process of teaching tennis.
    We have to understand that in the case of golf we have metrics A,B,C ...
    The goal is to increase A,B,C ...
    The question is like that: we cannot do the same in tennis because increasing power numbers
    Will send balls out
     
    Last edited: Mar 31, 2014
    #40
  41. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    IMHO, Julian, we should not put too much emphasis on the power and speed aspect.

    I think the upcoming features of swing path and position of impact on the racket (like Babolat does) are more helpful. (Though to be honest, not quite sure how the point of impact is determined with a sensor on the handle butt of the racket....)

    There are some aspects to be worked on: some of my high forehands are interpreted as serves, even though my height is a parameter specified for the sensor... Also, the spin statistics seem dicey to me... As far as power is concerned, helpful to see if one is generating more power over time and to see how the power holds up during a match/training session...
     
    #41
  42. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    I think the Zepp sensor is still a work in progress. I had played an hour this morning, 30 minutes training, 30 minutes match play. I hit at least 50 serves, and most of my backhands with topspin.

    The app counted only 8 serves (and 7 smashes - I hit only two), and said that 88% of my backhands were slice...

    Reported the incident to Zepp support, and hope to hear something soon...
     
    #42
  43. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Little use for coaching at this moment

    Phil,
    the basic part is missed in those conversations.
    I do not see how to use sensors at the current stage of development in my lessons.
    It is practically the only thing I am interested in.
     
    #43
  44. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Please keep me posted about feedback from Zepp

    Please keep me posted about a feedback from Zepp.
    By the way the sensor was released November 2013
     
    Last edited: Apr 1, 2014
    #44
  45. julian

    julian Hall of Fame

    Joined:
    Dec 20, 2006
    Messages:
    3,310
    Location:
    Bedford,Massachusetts,US
    Position of impact

    I do not know how can I use the info about a position of impact.
     
    #45
  46. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    Julian,

    Our goal for the future is to make this information accessible to anyone who needs or wants it. Of course, to be of use, we must make it as accurate as possible. To strive for anything less would betray our vision.

    Now having said that, there are of course players who may not have a need or a desire for it, and that's OK. For those that do want to know more about the details of their game, the capability will soon be there. Once we get deep into the testing phase, we can really evaluate just how accurate the data is when compared with the leading methods such as high speed video. This is something that's unavailable to most right now, and that's exactly what spurred us to develop this product.

    It goes without saying that no piece of hardware can ever replace a good (human) coach. :)
     
    #46
  47. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    SFeingold, any idea how mapping the impact position of the ball on the racket is done? (Babolat Play does it, Zepp plans it)

    Kind of baffling, since the sensor is in the butt of the racket, so all measurements and figures are relative to this point of measurement....
     
    #47
  48. SFeingold

    SFeingold New User

    Joined:
    Dec 7, 2013
    Messages:
    12
    I would wager that it has to do primarily with the vibration signature that is detected.

    So, if you were to look at a vibration analysis (such as a fast-Fourier transform) of a particular set of data (vibration after a stroke), you would see that different vibration frequencies are prevalent in different proportions depending on where the ball impacts the racket. You can use this to then characterize different shots. Although different tensions and string types will change this data, the main "signatures" that you're looking for should remain similar. 2 symmetrical shots about the centerline of the racket (left and right sides of the racket) may have very similar vibration characteristics and here the gyroscope can assist. There will invariably be just a little bit of rotation of the racket when the ball is hit. So, if the ball hits one side of the string bed, the racket will twist in that direction. Same for the other side.

    Babolat has an inherent advantage so far, because they are only concerned with one single racket and things are much more consistent. Once you start adding multiple rackets into the mix, it gets more difficult.
     
    #48
  49. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    [​IMG]

    This is the Gartner Group New Technology Hype cycle which certainly applies to all of these new sensor-based products. Obviously we are in the inflated expectations phase. Some of us, me, for example, have already progressed to the "trough of disillusionment", and await the next phase...
     
    #49
  50. gzhpcu

    gzhpcu Professional

    Joined:
    Apr 16, 2005
    Messages:
    1,019
    Location:
    Switzerland
    Thanks for your reply SFeingold, the torsion from the gyroscope will help. But doesn't firmness of the grip also play a role?
     
    #50

Share This Page