Coaching and consulting are well-worn career paths for retired athletes who still want to contribute to their sport — but champion Olympic figure skater Craig Buntin had different ideas.
He went into tech and expected his venture would one day branch out into software for self-driving cars.
Instead, figure skating’s reliance on technical judges to score figure skaters based on whether or not they completed rotations or other moves inspired him to go back to the ice as an analyst.
“That’s actually what started us moving toward sports,” Buntin says. He’s now the co-founder of Sportlogiq, a Montreal-based sports analytics company that uses data to help coaches, teams — and even entire sports leagues.
Buntin spoke to the Star about how top-level coaches use Sportlogiq’s data, why AI insights can’t tell you everything about how athletes perform, and how his company might have made his life easier as an Olympian:
I recall reading that you never expected to work in sports again after retiring as an Olympic athlete. Is that the case?
It was funny. As a kid, I was building computers and making websites for people to support my skating. So I always kind of enjoyed technology and enjoyed software, and I was really excited when I retired to really dive in. I didn’t really expect to make my way back to sports at all. When Mehrsan Javan, our co-founder, and I got together, we were originally trying to build tech that could help cars drive themselves. It was at some point, not long after that, we realized that we needed customers — and we knew a lot about sports and not a whole lot about cars.
How does the process of training AI for sports work? I imagine it to be somewhat like training an AI for autonomous cars — strapping cameras to it and driving around.
That’s exactly it. We’ve designed these algorithms to learn in the same way that a human brain learns. Let’s say someone is born into the world and they’ve never seen anything. All of a sudden, their eyes and ears work, and they say — teach me. And you were to show them three or four shots from the point. That person would only ever know that a shot came from the point and that’s how to make a shot. They would never understand the dynamics of a breakaway or putting players around the net, or trying to deflect shots. You can only really set the algorithm to truly understand what a shot is if it’s seen every different possible way that a shot can be done.
Have you ever found insights through Sportlogiq’s training that a human simply couldn’t have come up with?
That’s a really cool question. What we’ve found is that coaches — let’s talk NHL coaches — are really smart. When we started, we decided we were going to build these tools that will advise teams on how to play or who to trade. For the most part, the teams said: “Don’t tell us what to do. We know how to do this.” What we ended up doing is building tools that allow them to do their jobs better and more efficiently.
When we gave insights to, say, a head coach, we’re not really teaching them anything new about the philosophy or framework for how they think about the game. It’s pretty much matched with what the data says. But what the data does is like taking a top coach and allowing them to watch 5,000 games at a time. It allows you to pull insights out of a broader set of games.
So if you’re looking for a defensive line that moves the puck really well through the defensive zone, for example, you’ll be able to easily see any defensive line in the league that does exactly what the coach is looking for.
Sportlogiq is working on analytics for hockey, football, soccer and lacrosse. Are there any sports you’re interested in breaking into that you haven’t yet?
I don’t think I’ve ever thought about that question. When we started, neither Mehrsan nor I had really watched much hockey or soccer. The reason we chose to move into hockey and soccer and football was purely objective: we looked at the size of the market, we looked at the competitiveness of the technologies and products, and we just made a calculated decision to get into those sports.
I think hockey, soccer, and football are where we see Sportlogiq staying in the short to medium term. Baseball is probably one sport we’ll likely be getting to scale very soon, but that’s because the technology is really well-suited for baseball.
Can you tell me a bit about the work Sportlogiq does with fantasy sports leagues?
It started as just an internal thing. We’ve got pretty competitive fantasy leagues (laughs). We’ve kind of explored the possibility of opening up some of our data to fantasy leagues. We don’t have much of an open product for everybody yet. But look, if there’s any company out there reading this — we’d love to work with you!
If Sportlogiq had existed when you were training for the Olympics, do you think it would have changed your performance?
I’ve gotta tell you a story in order to answer that question.
In 2007, my figure skating partner and I were ranked fourth or fifth in the world overall in the Grand Prix circuit. The Olympics were three years away, in Vancouver. My partner decided that she was falling out of love with skating and she retired. So, I found myself three years away from the Olympics without a partner — and you don’t just pick up a partner and become one of the top pairs in the world. It doesn’t happen.
I drove about 3,000 kilometres across North America and went into local stadiums, got on the phone — anywhere — looking for a girl who’d have been the right match for me, who’d have had the right technical capabilities, and was willing to work with me. I ended up finding Meagan Duhamel. She ended up moving to Montreal the weekend after we tried out. Within a month, we were the sixth-ranked figure skating pair in the world.
That would have been a much easier process if I had been able to just simply look at the data from 1,000 different skaters and narrow down my search to one that would have been the right match.
Sportlogiq is geared toward teams or franchises, but what do athletes themselves say about the insights you gather?
It really depends on the athlete. There are certain players who would look at this data and it would throw off their game. They’d think about focusing on this metric or that metric, and they’d overthink what should otherwise be them feeling out how they should be playing the game. We’ve tried to keep that in mind as we deploy products. Sometimes, the best way to communicate with athletes is not just by handing them data. It’s by handing the insights to their trainers and coaches.
There are a handful of athletes who just take the data and run with it. I probably would have been one of them. But the vast majority are going to get the value out of it from actually working with the coaches who know them and know how they need to ingest the information.
This may sound like a philosophical question — given all the analysis you do, is there anything qualitative left in sports that you can’t or don’t expect you’ll be able to measure?
Why someone is doing something will always be philosophical. We can measure something that has happened, and we can talk about the statistical probability of it happening again, or the likelihood of a certain play happening. I can’t honestly look you in the eye and tell you why. It is purely speculative. I am no more ahead than any average fan debating something at a bar.
Sportlogiq analyses a lot of different sports. Are there any common markers of great players that you find in the data?
I don’t think you’re going to like this answer. Grit. It’s not a data question that comes up. The data simply tells us how that athlete is doing. When you’ve got an athlete who goes through injury or through major tragedy in their life and comes back stronger — the data can just tell you what’s happening. It can’t tell you why.
The interview has been edited and condensed.