It is now commonplace for artificial intelligence to defeat a professional Go or shogi player, so why not take AI to soccer?
J.League team Sagan Tosu has been developing an analysis system that assigns point values to each player's movements using AI. In collaboration with a developer, the system, which can quickly process large amounts of information and make judgments and analyses, is being tested by the club's youth teams.
During a soccer match, 22 players are almost constantly moving around the field. As it is difficult to read the developments of the game, unlike in Go and shogi where patterns form, the AI uses video of a match to determine the good and bad plays.
To establish the analysis system, AI developer LIGHTz Inc., headquartered in Tsukuba, Ibaraki Prefecture, asked coaches about the movements and roles required of players depending on the situation. For instance, it took the holding of weekly, two-hour sessions over five months just to cover what are good plays during the build up to turn defense into attack.
The AI system, powered by the expertise of the coaching staff, analyzes data generated from match footage to evaluate players. Even if players do not touch the ball, they are highly evaluated if the AI recognizes them as keeping an opposing defender occupied or effectively using space.
The club aims to apply the system to the top team from November.
"Our accumulated knowledge has been incorporated into this analysis system," said Yuki Seki of LIGHTz who was involved in development. "It has to process a ton of data, but that's why it's important to get AI to do the advanced calculations."
Analytics in baseball has led to the theory behind the so-called fly-ball revolution where hitting the ball at certain angles up in the air is more likely to generate hits, rather than the conventional approach of staying on top of the ball.
With the introduction of AI, playing styles and formations that have long been considered unusual in the soccer world might become mainstream.
Read more from The Japan News at https://japannews.yomiuri.co.jp/
