Sensor Accuracy Challenge Part 1 – Groundstrokes, Volleys and Serves
To see if we can get some sort of handle on the accuracy of the products available in a fairly young product market – tennis sensors.
Our starting point is the the simplest of accuracy tests. Can these little devices accurately detect the difference between forehands and backhands? What about serves and volleys?
In part 2 we’ll delve a little deeper into spin. Racket speed and ball speed isn’t part of this series at this stage, largely because the Sony Smart Sensor is the only one that gives a measure for both, while the other 3 either provide one or other. The Heat Map is a little more difficult to measure. We’ve done our own testing which can be seen here, but we’re not including the Heat Maps in this 2 part series.
And so to what should be the minimum standard in our accuracy test: do the sensors pick up groundstrokes accurately and are they all detecting a serve when a serve is the shot played? Let’s face it, if they’re not picking up that sort of data then we might as well throw the rest of it away because it will distort all the other data to the point of it being useless.
The Zepp, Sony Smart Tennis Sensor and Babolat Play score the highest here. Those 3 sensors picked up scores in excess of 98% accuracy and frequently 100% in a session. The Qlipp on the other hand, read forehands as backhands and vice versa a little too often for our liking. The best score the Qlipp had in a session was 85% but was as low as 70%. At that rate the data is next to useless as there’s too much distortion to place any value on the overall numbers.
The Sony STS is the only sensor to pick up any sort of decent score when volleying. The Qlipp and Babolat Play make an effort but all too frequently a volley is picked up as a groundstroke. The Qlipp still had the forehand versus backhand problem as well. The Zepp makes no attempt to read volleys at all.
While not as high as 98% accuracy on volley detection, the Sony still gets a tick from us. We have found that the Sony picks up volleys in excess of 90% of the time when a short back swing and follow through is used. The same can’t be said of the Qlipp or Babolat Play. Both devices still picked up far too many groundstrokes no matter how short the swing.
This is another reason why a shot-by-shot drill-down plus video analysis is high on our sensor wish-list. Without shot-by-shot drill-down, distorted data is displayed within a sensor app without the player knowing which pieces of the data were right and wrong.
The Zepp sensor comes out on top in the ability to read serves. We haven’t been able to fault it. Around 1 in 20 serves register as smashes with the Sony Tennis Sensor and the Babolat Play racket. The Zepp makes no attempt to read smashes while both Sony and Babolat Play do. The latter two still get our tick of a approval because a misread of a serve as a smash isn’t distorting our groundstroke data for a session.
The Qlipp, unfortunately, registered far too many groundstrokes as serves in our testing. Too mush data distortion means unusable displays within the applications.
It would be difficult to find a shot in tennis more difficult to read for a sensor than a smash. The motion, including follow-through, is so similar to serving. The two sensors that at least make an attempt to register a smash are the Sony sensor and the Babolat Play. In our testing, neither registered enough smashes to warrant us giving a big green tick.
The Zepp and Qlipp don’t try to register smashes. We have trouble understanding why no attempt has been made to register volleys, but smashes are different, largely because of the follow-through mimicking that of a serve, whereas a volley is far more abbreviated.
Before you throw the baby out with the bath water, there is a way to use the Babolat Play or Sony Smart Sensor to practice your smash. Details can be found here.
Sensor Accuracy Conclusion
The three most common swings used by a tennis player are forehands, backhands and serves. Sony Smart Sensor, Babolat Play and Zepp all distinguish between these different swingswell. It would seem that the Qlipp still has work to do on the most basic of tennis sensor algorithms.
As we get into the abbreviated, but similar, swing of the volleys and the similarity of the smash to the serve, all the sensors start to struggle. Some more than others.
The Sony Smart Sensor gets our top marks for basic accuracy, followed by the Babolat Play. The Zepp comes in third but makes no attempt to measure anything other than groundstrokes and serves. The Qlipp gets points for making the attempt but comes in last place because too much of the data is skewed by inaccurate reading of swings.