High-tech Data Analysis Coming to Aid of Israeli Judokas at Olympics

'It’s another screw in an entire system, but it gave us a full picture of the matches and rivals,' says men’s team coach Oren Smadja.

Send in e-mailSend in e-mail
Send in e-mailSend in e-mail
Roy Meyer of Netherlands and Or Sasson of Israel compete at Rio Olympics, August 12, 2016.
Roy Meyer of Netherlands and Or Sasson of Israel compete at Rio Olympics, August 12, 2016.Credit: Toru Hanai, Reuters

Behind the successes of Israel’s judo team at the Rio Olympics is an interesting venture that could be described as “startup nation meets training mat.” It was only in April that Aviv Hatzir, an entrepreneur and former judoka himself, started Boost to develop a virtual analyst for athletes.

The system was supposed to only be marketed later this year, but as part of the preparations for Rio and with a push from the Israel Judo Association and the Israel Olympic Committee, development was expedited for the benefit of the Israelis competing in Rio.

Boost is developing an analytics tool in the form of a chat-bot, which allows athletes and their trainers to ask questions about the athlete himself, his training or his rivals in an upcoming competition. The system, Hatzir explains, analyzes the question using artificial intelligence and gives the questioner an immediate response. In cases where an answer cannot be automatically provided because it requires more research, the question is referred to a human analyst, who gives an answer tailored to the question, which helps improve the system’s abilities in the future.

Boost joins a wave of startups that are exploiting the big-data revolution to develop analytic tools for all sorts of organizations. The sports world – particularly basketball and soccer teams – is a heavy user of tools to analyze various data. NBA teams, for example, are known to have teams of analysts who process all sorts of data gathered from devices and sensors on the court.

To start, Boost developed an engine that collected all the information that exists on the web about every rival judoka, starting from available technical information like physical characteristics, achievements and various statistics, and then analyzing all the matches the rival was in, based on thousands of videos that were coded automatically and manually. The next stage had the system process all the information using algorithms that were able to produce an exact profile of every rival athlete, with an emphasis on predicting future trends and identifying his strong points and potential weak points.

“It’s another screw in an entire system, but it gave us a full picture of the matches and rivals,” said Oren Smadja, the Israeli Olympic medalist in judo who now coaches the men’s national team. “When I competed in the Barcelona Olympics in 1992, we shlepped videotapes to study our rivals. Here it was all ready and it’s presented in an accessible fashion that allows us to ask questions at any time. You cannot cope with so much information without this tool, and we used it a lot. It helped us and proved itself, and we will continue to use it in the future.”

Hatzir explains that before Or Sasson’s match with Egyptian Islam al-Shahabi, which caused a media storm when the latter refused to shake Sasson’s hand, the professional staff believed that Shahabi would open the match with blows, rather than with classic judo moves. To better understand how to contend with this, the system was asked several questions; it identified the mistakes Shahabi tended to make in such combat and was able to give insights. Before the match with undefeated champion Teddy Riner of France, the system noted that when one grabs Riner in a certain way on his left lapel, it’s hard for him to fight at his best.

“It could be that this is information that coaches know, because he’s a known athlete, but we make the information and insights available for many other athletes,” said Hatzir. “We didn’t invent judo match analysis, but we streamlined and simplified the process.”

Click the alert icon to follow topics: