Waterloo researchers and Stathletes push the boundaries of hockey analytics

A partnership between the Vision Imaging and Processing (VIP) Research Group at the University of Waterloo and Stathletes - an Ontario-based professional hockey performance data and analytics company - is advancing research and accelerating the company's product roadmap.

At the heart of the collaboration is a grand challenge in computer vision: extracting reliable, event-level information from a sport played at blistering speed captured by cameras that are constantly moving.

"Hockey is a very fast sport," says Dr. David Clausi, director of VIP and professor in the Department of Systems Design Engineering. "We're trying to figure out where the players are, where they're moving to, what their actions are and what they're doing as a team." To do that, the team must estimate each player's X and Y location, infer the puck's position and cope with boards, bodies and motion blur that routinely hide key objects from view.

Even before artificial intelligence (AI) became mainstream, the VIP Research Group was applying it to introduce technical advances that adapt pose - a computer vision concept for understanding the skeleton of a body - to meet the unique challenges of hockey's fast-moving, cluttered visual environment.

Photo credit: Kseniia Buzko. Images depict a computer vision concept for understanding the skeleton of a body, also known as "pose."

Its major contribution extends conventional pose models beyond the player's skeleton to also locate and track the stick - an object thin, fast-moving and often partially invisible in broadcast feeds.

"If we have a good understanding of the stick, we have a really good understanding of what the player is doing," Clausi explains. "We've taken existing pose models and extended them to understanding the stick being held by the player at the same time." The richer pose-and-stick representation improves action recognition that distinguishes skating, shooting or protecting the puck, and strengthens sequence-level understanding of play.

Meghan Chayka, co-founder and co-CEO of Stathletes, says that the business case is clear: faster, more accurate data unlocks greater value and more use cases for professional teams and partners across the hockey ecosystem. "The more real-time you can get, the more valuable the data is," she says.

As the company sought to scale its output beyond the labour-intensive human-in-the-loop tagging that could take 12 to 15 hours for one game to reach the desired event-level detail, she discovered the VIP Group. Together, they explored how computer vision and machine learning could accelerate automation while expanding the depth and breadth of Stathletes' datasets. The collaboration has allowed the company to push the boundaries of research while operationalizing insights that measurably advance the business.

Clausi and Chayka describe the partnership as a win-win. Stathletes supplies the real-world data and problem framing that research needs, while VIP researchers provide innovative methods and people who can translate breakthroughs into production.

"Stathletes has been an essential partner … they support students with ideas, funding and direction," Clausi says. "They're also able to provide what we need in terms of training data. We build models around it and then ask, 'What other data do we need?'"

The result is a steady pipeline of Waterloo students gaining hands-on experience that helps them secure industry-leading roles.

"David has been such a good curator of people that are very inquisitive in computer vision and machine learning," she says. "Not only have we been able to work with the students, but we've also been able to hire them at Stathletes."

"Being connected to Waterloo and world-class research superpowers a lot of our work." - Chayka

In a field where innovations arrive weekly, Chayka emphasizes that success depends on people who can adapt quickly to further business goals, and Waterloo has a rich ecosystem of talent.

Looking ahead, both Clausi and Chayka see the collaboration accelerating exponentially. Chayka notes that Stathletes has recently hired Waterloo talent from the VIP to drive high-impact projects and deliver value to its clients.

For Waterloo, the partnership highlights the University's strengths in transformational research that translates into real-world solutions, economic impact and exceptional training for students who will become the next leaders of AI.