How is deep learning affecting sports? Posted on : Oct 18 - 2017

I don’t know about you, but I was not the most athletic kid growing up. It took me forever to make a jump shot well. When I started playing golf after college my short game was an absolute disaster. I always had a hard time visualising what I needed to do differently. Having a coach tell me what to do never seemed to do the trick.

Once, after a particularly awful few holes, a buddy of mine filmed me with his iPhone and just watching myself duff the ball made a huge impact. Deep learning has the potential to use simple data like video captured on a phone and turn it into a huge asset for sports – both players and fans.

Moneyball, the now legendary story of how the Oakland A’s used sabermetrics player analysis to outduel big money teams from New York, Boston, Los Angeles and elsewhere, is widely considered to have changed the way Major League Baseball franchises build their teams.

Few people expect teams of humans to be taking on squads of AI robots on the courts or in the field anytime soon, if ever. Yet deep learning, a subset of AI focused on pattern recognition and reinforcement learning, is already in use by researchers and technology startups across a variety of sports. Researchers are training their deep learning applications on large sets of data, and then deriving strategic insights into areas such as player capabilities, game patterns, and team tactics.

When deep learning becomes the game changer

Let’s take the example of America’s traditional pastime: baseball. New York University Professor Claudio Silva and MLB Advanced Media consultant Carlos Dietrich have developed a metrics engine that tracks each movement of every player and the ball throughout a game. The system takes in all this data and then identifies patterns that can help coaches manage players, plan strategy, and equips them with the ability to make predictions about the game. This may include where fielders should set up for certain hitters, or what specific pitchers like to throw based on the hitter, the count, or where they are in their pitch count.

In ice hockey, Toronto-based startup ICEBERG is using AI and computer vision to give teams a nuanced understanding of the data behind player positioning and activity. The firm collects video of a game and then turns it into millions of data points based on the movement of the players, the opponents, and the puck. View More