There we were, sitting down with Sportlogiq CEO Craig Buntin in his Mile-End office. Our eye balls followed a screen showing the players in this year’s Stanley Cup final between the Pittsburgh Penguins and the San Jose Sharks.
Buntin had us focus on a dozen or so of 3,100 distinct events in a hockey game that his computer vision software at Sportlogiq is tracking. Nearly half the teams in the National Hockey League (NHL) are paying customers for the data and the actionable analytics the company provides.
From a face-off loss to a loose puck recovery in the neutral zone, from an outlet pass to a pass reception to a “d-to-d pass,” we followed maybe 30 seconds of an NHL game in this step-by-step manner.
“If you watched hockey in Canada this year, you saw our stuff,” Buntin told MTLinTECH. “Whether it was the on-air talent talking about players, the pre-game show graphics or in-game stats, we were involved with almost every broadcast.”
Sportlogiq’s technology generates advanced hockey analytics using standard, single-camera game footage. The computer vision system uses a combination of player tracking and activity recognition algorithms to flag specific game events such as shots or passes, it timestamps them and records their x-y coordinates on the ice. Once the raw data is collected, machine learning and pattern recognition techniques are utilized to extract meaningful insights, allowing broadcasters, teams, analysts and fans see and understand the game in a way previously not possible.
In other words, the forefront of advanced sports analytics in NHL hockey is (probably) happening right here in the Mile-End. Mark Cuban, the infamous owner of the Dallas Mavericks, has already invested in the startup (which began from a simple cold call made by Buntin one day. Perk up, startup founders).
For those not familiar with the NHL hockey world, front-office execs across the league have slowly warmed up to “advanced analytics” over the past three years. They want to know the statistical tendencies of their own team and opposing teams in order to adjust strategy and make decisions that can have a greater outcome on any given game.
It started as a fad when teams like the Toronto Maple Leafs hired Kyle Dubas as an assistant GM because of his advanced stats knowledge, or when Edmonton hired Tyler Dellow. Then, it seemed, everyone was bringing on an advanced stats nerd.
The thing is, these teams are so hell-bent on winning that they don’t want anyone else knowing what they’re looking at. For a while before the Carolina Hurricanes made it official, SB Nation’s Eric Tulsky was working for a team that wouldn’t let him say who they were (probably the Hurricanes). And so with Sportlogiq, Buntin refused tell us who his clients in the NHL were. Not only that, the majority of the company don’t know which teams are currently clients.
Thus, one can imagine how much money these NHL teams are willing to shell out for the best analytics. Buntin is convinced he’s the guy who is delivering the best product.
“Anything you can do with professional sports data, right at the base of that is data collection, and that’s what we’re doing better than anyone else,” said Buntin. “Essentially everything you could possibly annotate in a game, we have. The question then becomes, ‘What can you do with that?’ So you can understand he magnitude of information we have.”
“My life has been a series of swing-for-the-fence events,” the 36-year-old CEO from Kelowna, British Columbia told us.
He bought a one-way ticket to Montreal as an 18-year-old to train with the best pairs figure skating coaches. He was “arguably one of the bottom skaters in the country.” Still, he trusted himself and within a year was a junior Canadian champion. Buntin went on to represent Canada at the 2006 Olympics in Turin, Italy, and has won three Canadian championships.
After retiring from figure skating in 2010, he put his own cash into a coffee business that was eventually acquired by its distributor. Still, he realized that business was really hard and he wanted to gain formal education. Despite having just a high school education and being out of the classroom for over a decade, McGill University made an exception and allowed him into its MBA program.
“I did an independent study, took a semester to study everything that I potentially wanted to do and came out of school with a business plan, and that’s what turned into Sportlogiq,” he said.
Sportlogiq’s suite of data tracks individual players and teams as a whole through 140 different things (“events'”) they can do in a hockey game. Each player is given a success or fail rating on each play based on puck possession. Variables like man strength (five-on-five, powerplay, etc.) and more are taken into account.
The software is not 100 per cent automated yet, but Sportlogiq is close to getting there. Soon it won’t even need the help of a human running a quality control check.
“When you think of machine learning and where neural networks are, these things are going to get just as good as humans. I personally believe that over the next couple years this will be as strong as any good human being can be,” said Buntin.
In terms of distribution, Sportlogiq can sell its analytics to professional sports teams, TV broadcasters and digital broadcasters, video game publishers or as a tool to be used as an aid in sports betting.
For now the startup wants to completely dominate the NHL market. It will wait before it expands into other sports like basketball, soccer, baseball and football. It’s currently scaling out across five other hockey leagues so NHL team clients will have access to the best information as they seek to make deals between leagues.
In using the data to aid sports betting, Buntin was adamant that Sportlogiq will never become a betting company. But he sees a “huge opportunity” in providing better data and analysis to people who want to know how certain players and teams will do in certain games.
In working on a sports prediction patent, Buntin said “we realized that we could actually predict game outcomes better than anybody else. We’re actually predicting the winners of games with higher accuracy than can be done with NHL data.”
Over the next year, the team will look to hire more PhDs in computer vision, raise a series A round of funding and, ideally, take over the NHL’s data market. We’ll be excited to see what happens next.