Artificial intelligence is redefining the future of sports analytics, and nowhere is that more evident than in hockey. Researchers at the University of Waterloo have developed two advanced AI systems capable of tracking puck and player movements with unprecedented accuracy—using only standard video footage. This innovation could dramatically lower the barrier to professional-level analytics, making data-driven insights accessible to amateur teams and smaller organizations.
At the core of this research are two systems: Puck Localization Using Contextual Cues (PLUCC) and SportMamba. Both represent major leaps in computer vision and automated sports tracking. Traditionally, tracking hockey action has been difficult due to fast puck movement, frequent obstructions, and motion blur in broadcast feeds. But the Waterloo team’s AI models can now infer game dynamics with a human-like understanding of the sport—essentially giving computers “game sense.”
PLUCC, led by graduate researcher Liam Salass, takes an innovative approach by observing players’ body positions and eye direction to predict where the puck is, even when it’s hidden from view. “Finding the puck in broadcast video is one of the toughest problems in sports vision,” Salass said. “Seeing our system accurately predict its location using contextual cues was incredibly rewarding.” The system improved puck detection accuracy by 12% and reduced localization errors by over 25% compared to existing solutions. For smaller teams or community-level hockey leagues, this offers a low-cost alternative to expensive systems like Hawk-Eye.
Meanwhile, SportMamba tackles the challenge of tracking multiple players in fast-paced sports. Developed under the guidance of Professors David Clausi and John Zelek, this AI-based framework dynamically predicts movement trajectories, even when players are obscured or when camera angles shift rapidly. Tested across multiple sports—hockey, basketball, and soccer—SportMamba delivered up to 18% higher tracking accuracy and allowed for real-time analysis without sensor-based systems.
As Dr. Clausi noted, “These improvements in detection accuracy could transform how coaches, teams, and broadcasters analyze game dynamics.” Dr. Zelek added, “It’s much more difficult to track players in a scrum or in front of the net, but SportMamba can tell us exactly who deflected the puck or scored.” Together, their systems form the foundation of the Vision and Image Processing (VIP) Lab’s mission—to make advanced analytics practical and affordable for everyone involved in the game.
Both research papers—“Ice Hockey Puck Localization Using Contextual Cues” and “SportMamba: Adaptive Non-Linear Multi-Object Tracking with State Space Models for Team Sports”—were recently presented at the 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, showcasing Waterloo’s leadership in the intersection of AI, sports, and engineering innovation.
Conclusion: The University of Waterloo’s breakthroughs mark a new era for AI-driven sports analytics. With systems like PLUCC and SportMamba, even teams with limited budgets can harness the power of artificial intelligence to make smarter, faster, and more strategic decisions on and off the ice. The days of million-dollar tracking setups may soon be over—replaced by accessible, intelligent tools built for the future of hockey.
