How artificial intelligence is changing hockey
Data has revolutionized the Canadian pastime
In a classic startup setting — in a former textile plant four miles from where the first hockey match was played a century and a half ago — a group of high-tech computer engineers is changing Canada’s most revered pastime.
There — in sterile cubicles amid lots of wood and windows, with a jelly-bean dispenser and the inevitable dog — they examine the 4,000 motions they detect players make in the course of each 60-minute game. The result is millions of data points unavailable to fans in the stands, but indispensable for coaches and, ultimately, players.
The work being done here is changing the world of sport. But its real significance is as a measure of how artificial intelligence is changing our world.
The conventional wisdom is that metrics have taken over the elegant game of baseball, but the lesser-known truth is that the rougher game of hockey is being shaped just as decisively. Indeed, analytics is the new power play of a game that began modestly, with nine players on a team batting around a wooden puck.
The game has changed since March 1875: forward passes now are permitted and boards, glass and netting now surround the playing surface. But these new insights — think of them as the result of the collision of techies and goalies — that are being produced through artificial intelligence here at a firm called Sportlogiq are transforming hockey at a ferocious speed.
“The game now is being played differently,” says Craig Buntin, a former Canadian Olympic pairs skater and co-founder of the company. “There’s always been a disconnect between analytics and coaches. We’ve made millions of data points suddenly useful for coaches. Artificial intelligence has given teams eyes where they didn’t have them.”
To the naked eye, or at least to mine, a recent game between the Montreal Canadiens and the Vegas Golden Knights was a push-andshove match that the Habs won, 54, in overtime, simple as that. But the next morning, in the Mile End section of Montreal, Mr. Buntin pulled up an image of the game’s passes to the slot in the neutral zone. In that spaghetti image was the future of hockey, available in time for coffee and Montreal’s famous wood-fired bagels.
No traditional statistic reveals nearly as much information. “You win games scoring more goals than are scored against you,” says Mr. Buntin. “You need to know how you score those goals. So you need to know where the high-probability shots came from and how they are generated.”
The data that Sportlogiq conjured from examining every motion of every player 30 times a second revealed the Canadiens had higherquality power-play shots than the Knights, but the Vegas players got a lot more even-strength shots when the player wasn’t being pursued and had much more accuracy on outside shots. They also showed Montreal had an early advantage with high-quality shots but by the third period, Vegas was dominant in that category.
Who knew? And who cares? The Pittsburgh Penguins, for example, care about knowing such things. They made a trade for defenseman Ian Cole in March 2015 — “one of the most intelligent trades I’ve ever seen,” says Mr. Buntin — based in large measure because of data points.
“He was a really strong puckdriving player,” says Mr. Buntin. “He could move the puck forward from the defensive zone to the offensive zone really well. He wasn’t going to score a lot but he could get the puck to the players who could. That wasn’t showing up in any other metric.”
The Columbus Blue Jackets care, too, especially since Sportlogiq last spring helped them pull off a dramatic upset playoff sweep over Tampa Bay by understanding that the Lightning was vulnerable off fore-checks in its defensive zone.
Sportlogiq did that by noting the location of every player at every moment and producing a full three-dimensional reconstruction of their bodies and their sticks. “We’ve built a box that can see, understand and describe the game,” says Mr. Buntin. “That can help teams find the specific-style players they are looking for, identify an opponent’s strengths and weaknesses, and see how certain lines are being beaten.”
The key is a combination of artificial intelligence and the intelligence of Christopher Boucher, a former goalie and onetime hockeystatistics blogger who spent a decade in his basement watching, and analyzing, hockey videos. He was literally a game-changing Sportlogiq acquisition.
As a result of the insights of Mr. Boucher and his Montreal team, hockey teams around the globe — all but a handful in the National Hockey League, the American Hockey League and the Swedish Hockey League — are harnessing AI.
“Analytics and data have become a part of how we evaluate players and how we evaluate performance,” said Penguins president David Morehouse. “It’s an important element of our tool box. We’ve made big moves because of it.”
But what of the romance of the sport, and of its tradition of leaning on hard-earned hockey intelligence from hard-bitten hockey men who played hard in more primitive circumstances? Is that obsolete?
“There’s still something intangible about a team’s ability to assess, react and improve,” says Mr. Buntin, who retains a strain of romance about the game. “We look to sport to see human beings doing seemingly impossible things. We help them do that. The [old-fashioned hockey hand] doesn’t see what we reveal. But he’s going to see players excel at what they do well. For me it’s goose-bump territory.”
Then again, there’s no romance in sport to compare with hoisting the Stanley Cup. Emotional intelligence is one thing, but in the language of hockey, artificial intelligence is emerging as a game-winning assist. It’s coming to every aspect of our lives.