How did the Columbus Blue Jackets sweep the Tampa Bay Lightning in the first round, pulling off one of the biggest upsets in NHL playoff history? Artificial intelligence may be partly responsible.
A post-series analysis by Montreal sports data analytics startup Sportlogiq Inc. revealed Columbus exploited one of Tampa Bay’s few weaknesses – allowing shots and goals off forechecks in its defensive zone. Columbus also limited Tampa Bay’s top-ranked ability to move the puck through the neutral zone.
Sportlogiq uses a form of AI called computer vision to derive deeper data-driven insights from hockey games than had been available before – including the forechecking stats that ranked Tampa Bay near the bottom of the league. One of its customers is Columbus, whose director of hockey operations Josh Flynn confirmed Sportlogiq stats “were used to prepare a preseries analysis for the team. The forecheck stats were part of that.”
In an era when analytics has become a bigger part of pro sports – and has seen the rise of a range of statistics companies including Opta and Stathletes – Sportlogiq has become a “need to have” source of insights for the NHL, said Steve Dryden, senior managing editor of hockey for The Sports Network.
All but three of the NHL’s teams use Sportlogiq – as do broadcasters TSN, Sportsnet, TVA, RDS and MSG – for data, insights and video snippets. The NHL is testing Sportlogiq technology for a puck and player tracking system, and the Swedish Hockey League uses Sportlogiq as its official data supplier. “The amount of data they aggregate … is tremendous,” said TSN director of scouting Craig Button.
The company has raised $30-million in financing, including backing from Mark Cuban, owner of the NBA’s Dallas Mavericks, and eight other pro sport teams owners. Sportlogiq is starting to offer similar analysis in other sports including soccer.
“I think there is enough of an opportunity in sports that this could be a multibillion-dollar opportunity,” chief executive Craig Buntin said.
In just five years, Sportlogiq has had an impact on the way pro hockey is played, strategized, scouted and analyzed, say several users interviewed by The Globe. Chris Boucher, vice-president, sports development, analytics and hockey operations, said teams are taking more effort to get the puck to the high danger area in front of the opponent’s net and taking more of their shots from there in response to Sportologiq data showing this led to more goals.
Insights from Sportlogiq’s analytics have found their way into pregame shows including teams’ relative strength in the neutral zone. It’s influencing trade decisions and how voters for NHL awards cast their ballots.
“It’s going to change the way the game is studied by people in the business and fans and ultimately it could change the way the game is played,” Mr. Flynn said. “It’s a good check on what you see with your eyes and helps you dig deeper.”
Mr. Buntin is a former Canadian pairs figure skater and Olympian who set out to build a self-driving car company after earning his MBA at McGill University in 2013. Within six months, “I realized I knew nothing about cars and I knew a lot about sports.”
Seed financier Helge Seetzen met Mr. Buntin while judging a business plan competition and invited him to join his Montreal incubator, TandemLaunch, after watching a video of him at a 2008 figure-skating competition. Despite gashing his hand on partner Meaghan Duhamel’s skate, Mr. Buntin returned to the ice and they performed a second-best long program as blood seeped through his bandages. “I thought, ‘I want to work with this guy, he’s a machine,’ ” Mr. Seetzen said.
Mr. Buntin joined with chief technology officer Mehrsan Javan – whose PhD thesis at McGill under the supervision of renowned computer vision professor Martin Levine, now an adviser to the company, is at the core of the company’s patented technology – in 2014 to launch Sportlogiq. Mr. Buntin initially thought figure skating was a good market until Mr. Seetzen convinced him to go after pro sports, where the money was. They started with hockey.
The timing was ideal. NHL teams were hiring analytics specialists to give them an edge. But to make the system work, Sportlogiq needed a huge set of “training data”: broadcast video feeds of games broken down into their individual components, with each play tagged by humans familiar with the sport so the machine could correctly identify plays after watching enough of them.
That’s when they found Mr. Boucher. The one-time Quebec Junior A goalie was a manager with courier company Purolator who moonlighted as a hockey stats junkie, watching recorded games 40 hours a week and systematically documenting each play. (The married father of two says “I have a really understanding wife.”) He’d done that since 1991, using a VCR, pencil, paper and calculator early on.
He wanted to understand how plays came together and created scoring chances, but there were few stats available. So he invented his own, breaking down hockey games like no one had before – akin to what Bill James did for baseball.
“I had a hunger to know why things were happening,” he said.
Metrics he created include “scoring chance generating plays,” measuring a player’s ability to get the puck in front of an opponent’s net, and “possession driving plays,” quantifying the ability to move the puck up ice. Mr. Boucher consulted on the side to NHL teams and wrote a blog, which is how Mr. Buntin found him.
Mr. Boucher was blown away by how computers gathered stats he’d gathered manually for years, at a speed and scale he’d never imagined. He became Sportlogiq’s fourth employee in 2015, giving up security and a pension and taking a pay cut. Mr. Boucher went from managing 35 drivers to overseeing more than half of Sportlogiq’s 80 employees, including many PhDs.
For four years the team has trained Sportlogiq’s algorithms to watch hockey to the point where the system is 99.5 per cent accurate, he says. But machines don’t do all the work – employees still watch the games alongside the machines, providing “quality assurance” to make up for some gaps in the technology. Teams and broadcasters get reports often minutes after periods end.
“I realize how lucky I am to be doing what I’m doing and having the ability and the potential impact I’m having on the game,” Mr. Boucher said. “This is pretty overwhelming.”