Like many European countries, France takes its soccer — or football — very seriously. To further refine the game of top players, a French startup called Footbar created a smart device to measure players’ stats such as speed, stamina, and skill. The artificial intelligence used by the Meteor tracker determines the quality of passes, shots, dribbles, and tackles based only on the acceleration of the tracker worn on the player’s calf.
However, the AI was unable to tell whether some athletes “cheated” by wearing the tracker on their ankles, which provided better statistics than when wearing it on their calves. To solve the problem, they sought out an intern at MIT.
When math and computer science junior Benton Wilson applied and was accepted for an internship through MIT-France, part of the MIT International Science and Technology Initiatives (MISTI), he had only a few expectations: to do something data science-related, to practice his high school French, and “to gain some perspective for my worldview.”
He joined four French interns at Footbar, located within Paris’ Le Tremplin, an incubator for sports-related companies. At first, he was "un peu nerveux" about using his French skills in a professional setting, but his high school soccer skills came in handy when he tried out the quarter-sized device himself.
His first task was to create algorithms that would determine where a player was wearing the tracker.
Players can upload their stats to their computer, view their progress, and compare their results online against teammates, as well as Meteor-wearing clients all over the world. “At the end of a game, you can get your stats for the game, and then you can also make a profile and see how you change over time,” explains Wilson.
But results are skewed depending on where players wear the tracker. Footbar decided that Wilson needed to develop an algorithm that would “penalize” the ankle-wearing players.
“It is a pretty complicated task for someone discovering the data,” says Wilson’s supervisor at Footbar, data scientist Sébastien Benoit. “His first weeks were probably a bit complicated for him as he both had to get more familiar with both our stack (understand how our code works) and with the type of data we work with (time-series data from an accelerator).”
Over the summer, Wilson used his background in machine learning and Python, and picked up skills in GitHub, the database system Django, and signal processing, to work on the algorithm. Not only did Wilson solve the problem, he discovered some technical solutions that further impressed his supervisor.
“After showing him a few examples, Benton was able to build a model and train it on enough data to make it work well,” says Benoit. He says that Wilson’s model was 95 percent accurate, and is now being used by Footbar’s production department. “We can now successfully detect the smart guys who intentionally exploited this flaw. Thank you, Benton!”
Benoit was impressed enough to let Wilson work independently, which led to him spending his final two weeks of his internship solving a second problem that had vexed Footbar: to automatically detect which of four fitness tests was taken by an athlete: sprint; sprint down and back; running endurance test; and vertical jump test. “Some of it was difficult, such as detecting when jumps tests occur versus other kicks/jumps, but overall I just worked on trying different things,” Wilson says.
“This task [was] probably twice as difficult as the previous one, but Benton completed this quite well and pretty quickly,” Benoit says.
Wilson added that he appreciated working within a tight-knit community that began each day with standup meetings, ate lunch together, and gathered for soccer, cross-fit training, and jogs.
In his off-hours, Wilson shared an apartment with a fellow MIT-France intern who was researching environmental sustainability for another company. Wilson joined a local gym, shopped in local markets, watched old movies at the Latin Quarter’s Le Champo theater, attended the China vs. South Africa Women’s World Cup match, ventured to different parts of France, traveled to Barcelona and the Netherlands, and people-watched along the Seine River.
“My favorite areas were over near the Canal de St. Martin and La Villette, where there are a ton of restaurants and places to sit along the canals,” he recalls.
He was one of 45 MIT students who participated in MISTI’s MIT-France this past summer. Its internship program, founded in 2001 with a collaboration between the French Ministry of Foreign Affairs and MIT, provides opportunities for research and experience in French companies and labs.
While the logistics of studying abroad can be daunting for many students, Wilson and his peers received a lot of help. MISTI programs cover basic needs, including airfare, housing, and food; Wilson and several others were also co-sponsored by the MIT European Club. Additionally, students receive support with their visas and other paperwork. MIT-France provides a comprehensive 377-page student guidebook to living in France. Student internships are between three to six months; Wilson stayed for three.
“MISTI has provided me with a unique opportunity to immerse myself in a totally new environment,” says Wilson, who adds that his experience gave him the confidence to consider an international career after graduation.
The deadline for next summer's internships is Dec. 1 for priority applicants. To apply, please visit mitfrance.com.
