Researchers are using AI to revolutionize how hockey data is collected

Trying to collect advanced statistics from a hockey game is like herding cats. The players are everywhere and they’re moving really fast.

This is one of the reasons hockey has been slower to adopt the kind of analytics seen in other sports like baseball. It is very difficult for humans to keep track of everything happening on the ice at any given moment.

The problem inspired researchers at the University of Waterloo to use artificial intelligence to develop new ways to collect game data.

Dr. David Clausi is leading the team in the Department of Systems Design Engineering which includes professor Dr. John Zelek, research assistant professor Yuhao Chen and a team of graduate students.

“It works by taking a broadcast game video and feeding into the deep learning algorithms we have in order to produce an understanding of where the players are, where they’re moving and an idea of what they’re doing on the ice.” said Clausi.

A press release from UW said the tools accurately track players at a rate of 94.5 per cent. The tools identify teams at 97 per cent and identify individual players at 83 per cent.

The team is working in partnership with Stathletes, a hockey performance data and analytics company based in Ontario.

As the algorithms are refined, Clausi believes the tools can be used outside the sporting world.

Top Stories

Top Stories

Most Watched Today