× close
Frames from National Hockey League broadcast video clips.Credit: Athlete
Researchers at the University of Waterloo are now able to capture and analyze data from professional hockey games faster and more accurately than ever before, with the valuable help of artificial intelligence (AI) tools. It had a huge impact on business. The research paper “Player Tracking and Identification in Ice Hockey” was published in the journal. Expert system with applications.
The growing field of hockey analytics currently relies on manual analysis of game video footage. Professional hockey teams across sports, and the National Hockey League (NHL) in particular, rely on that information to make important decisions about players' careers.
“The goal of our research is to interpret hockey games through video more effectively and efficiently than humans,” said Dr. David Krausi, professor in the School of Systems Design and Engineering at the University of Waterloo. “It's impossible for one person to record everything that's happening in a game.”
Hockey players move fast in a non-linear manner and skate dynamically across the ice with short shifts. Aside from jersey numbers and last names, which don't always show up on camera, uniforms aren't a strong tool for identifying players, especially in fast-paced speed hockey. This makes it extremely difficult to manually track and analyze each player during a match and is prone to human error.
The AI tool developed by Clausi, University of Waterloo School of Systems Design Engineering Professor Dr. John Zelek, Research Assistant Professor Yuhao Chen, and a team of graduate students automates and improves player tracking analysis using deep learning techniques. .
The study was conducted in partnership with Stathletes, an Ontario-based professional hockey performance data and analytics company. The researchers examined video clips of NHL broadcasts frame by frame, manually annotating teams, players, and their movements on the ice. They ran this data through a deep learning neural network, teaching the system how to watch games, compile information, and generate accurate analysis and predictions.
When tested, the system's algorithms achieved high accuracy. It scored 94.5% for accurate player tracking, 97% for team identification, and 83% for individual player identification.
While the research team is working on improving the prototype, Stathletes is already using the system to annotate video footage of hockey games. The commercial potential extends beyond hockey. By retraining the components of the system, it can also be applied to other team sports such as soccer and field hockey.
“Our system can generate data for multiple purposes,” Zerek said. “Coaches can use this to develop game-winning strategies, team scouts can look for players, and statisticians can identify ways to give their team an extra edge on the rink or on the field.” has the potential to truly transform the business of sports.”
For more information:
Kanav Vats et al., Player Tracking and Identification in Ice Hockey, Expert system with applications (2022). DOI: 10.1016/j.eswa.2022.119250
