Like many other popular sports, tennis is increasingly defined by data. Cameras, Hawk-Eye ball-tracking systems and teams of analysts now scrutinise every moment of every game, recording statistics ranging from the number of aces and double faults to the amount of unforced errors, net approaches and break point conversions.
The scope of modern data capture and analysis is vast. At the Wimbledon 2017 championships, for example, IBM's Watson Discovery Service crunched 22 years of unstructured information. It analysed a colossal 53,713,514 data points, providing match stats and insights in real-time for commentators, coaches and spectators alike.
This data can be used in different ways. For spectators, it can deepen the immersion, delivering live stats, comparisons and historical perspective to enhance the fan experience. Before a ball has even been served, predictive analytics can identify key performance indicators that could decide a match. While post-game, a platform like IBM’s SlamTracker enables you to see where points were won (and lost) in extraordinary detail.
For coaches, real-time applications like SAP Tennis Analytics can provide much-needed information about player performance. For big data analytics doesn’t just help visualise what happened during a match, it offers compelling evidence as to ‘why’ it happened.
Consider Roger Federer’s triumph over Marin Cilic at the 2018 Australian Open Final. The 36-year-old Swiss had already beaten his opponent eight times in nine FedEx ATP Head2Head meetings. But it “was his stronger service consistency – 67 of 84 first-service points won and 32 of 55 second-service points won – that ensured he was able to remain in contention [at Melbourne Park]... Federer also hit 24 aces to Cilic’s 16, converting six of his 13 break point opportunities on the Croatian's serve.”
Federer wasn’t just the odd-on favourite with pundits, data predictions also forecast his victory. At Tennis Australia’s Game Insight Group (GIG), they modelled potential outcomes of the tournament using Elo career ratings.
“One of the most reliable ways to set our expectations of results in tennis are with player Elo ratings,” says Stephanie Kovalchik, Tennis Data Scientist at the Game Insight Group at Tennis Australia. “Tennis player Elo ratings measure a player’s strength at any particular time – accounting for a player’s past results and the strength of their opponents – and are one of the most predictive measures of what a player is likely to do in the future.”
The more data you have to work with, the more accurate any modelling can be.
Adjusting ratings for injury and hard court surface experience, GIG repeated its Australian Open simulation 5,000 times to determine how matches would unfold. The forecast gave Federer a 38.9% chance of winning, followed by Novak Djokovic (20%) and Rafael Nadal (8.4%).
The more data you have to work with, the more accurate any modelling can be. At least in theory. Like cricket, football and many other sports, there are variables that can’t typically be quantified – weather conditions, a player’s emotional state or level of fatigue, injury, crowd noise and so on. Any prediction is still a ‘best guess’, albeit one based on hard facts and figures rather than guided by gut instinct.
On the coaching side, big data analytics can have a huge impact on player performance. By tracking a variety of key metrics (from serve direction to shot placement), it’s easier to identify areas of weakness and opportunities for improvement.
“With a slight adjustment to his ball toss, Federer can take an opponent wide on the advantage side service,” suggests John Quartararo, a Senior Consultant for Essentia Analytics. “With a subtle shift to angle of his racquet head, his backhand slice will skid even lower off the ground on a grass surface… Small changes in behaviour or process can create a meaningful difference in results.”
No wonder then that several sports analytics firms have sprung up hoping to help players find meaning in the data. Golden Set Analytics, for example, offers players access to statistics and game analysis designed to help them improve their performance. It also offers a scouting service, providing data on all the top 150 ATP players over an average of 25 matches played.
“Our reports provide advanced statistical analysis on a player’s strengths and weaknesses during match play,” says the company, “including over 200 tables of statistical information and visuals on serves, returns, groundstrokes, and advanced serve analysis. Scouting and match-up reports are delivered during tournaments within hours after draw.”
SAP is another big name in this area. The company works with the Women’s Tennis Association (WTA) to provide in-match data via its SAP Tennis Analytics for Coaches, powered by SAP HANA software on Intel® Xeon® Scalable processors. Using the on-court coaching rule at WTA events, this big data approach gives coaches the ability to reference in-match data to help players adapt and adjust game strategy.
“The WTA introduced the on-court coaching rule in 2008,” explains Jenni Lewis, Global Sponsorships Technology Lead at SAP. “That gave SAP the opportunity to bring real-time data to players and coaches as they need it… And they need it as the match is happening, so the coach can go out during on-court coaching and share that information.”
Again, the more data you have to work with, the more accurate any modelling can be. Beyond stat-scribbling analysts, number-crunching Xeon processors and high-speed Hawk-Eye cameras, the next level of data capture lies with wearable technology and Internet of Things devices.
The Babolat Play Pure Drive tennis racket, for example, incorporates a suite of sensors into the handle that measure power, points of impact, plus the type and number of strokes. The Zepp Tennis 2, meanwhile, offers a simpler approach. It incorporates its movement tracking accelerometers and 3-axis gyroscopes into a Bluetooth device that will fit onto the end of any racket.
Taking wearable technology a step further, the TuringSense PIVOT system doesn’t just use one sensor, it employs several for greater accuracy. Each one features an accelerometer, gyroscope, and magnetometer to process up to 1,000 data samples per second, tracking extra metrics other single sensor solutions can’t, such as footwork and body positioning.
It all adds to the data pool, where the secrets to success (or at least incremental improvement) lie hidden among billions of ones and zeroes. In tennis, the margin between victory and defeat is often paper-thin and even a 1% performance boost might make a difference. With insights gleaned from big data analytics, players can strive to up their game and fans can appreciate their every move.
*Other names and brands may be claimed as the property of others.