The 2019 edition of the MIT Sloan Sports Analytics Conference is right around the corner so we wanted to take a look back at our SSAC Research Papers and Posters from over the years. This year, 3 of our papers were published—including 1 finalist—and we’re excited to share our work on face-offs, team-level pace, and automated highlights/lowlights with everyone. Read all about them below.
2019 Research Papers & Posters
Winning Isn’t Everything: A Contextual Analysis of Hockey Face-offs
Authors: Nick Czuzoj-Shulman, David Yu, Christopher Boucher, Luke Bornn, Mehrsan Javan
Presentation by Nick Czuzoj-Shulman on Friday, March 1st at 3:15pm in the Research Papers Room (311).
This paper takes a different approach to evaluating face-offs. Instead of looking at win percentages, the de facto measure of successful face-off takers for decades, it focuses on the game events following the face-off and how directionality, clean wins and player handedness, play a significant role in creating value. This demonstrates how not all face-off wins are made equal: some players consistently create post-face-off value through clean wins and by directing the puck to high-value areas of the ice. We propose an expected events face-off model, as well as a wins above expected model, that take into account the value added on a face-off by targeting the puck to specific areas on the ice in various contexts, and the impact this has on subsequent game events.
Playing Fast Not Loose: Evaluating Team-level Pace of Play in Ice Hockey Using Spatio-temporal Possession Data
Authors: David Yu, Christopher Boucher, Luke Bornn, Mehrsan Javan
This paper provides the first comprehensive study of pace in hockey, focusing on how teams and players impact pace in different regions of the ice, and the resultant effect on other aspects of the game.
Our study demonstrates how the spatio-temporal characteristics of pace change in different game contexts, and that these characteristics have a meaningful impact on many game events of interest, creating odd-man rushes and impacting shot quality.
Lastly, we highlight the considerable variability in how teams and players attack and defend against pace. Taken together, our results demonstrate that measures of team- and player-level pace derived from spatio-temporal data are informative metrics in hockey and should prove useful in other team sports.
Data-driven Lowlight and Highlight Reel Creation Based on Explainable Temporal Game Models
Authors: Evin Keane, Philippe Desaulniers, Luke Bornn, Mehrsan Javan
This paper describes a framework for automatic highlight reel extraction based on game event data. The framework can be applied to most team sports given the output of any in-game state valuation model, of which there are many. We incorporate many levers for a producer to fine-tune the reel, such as adjusting the valuation of events based on their impact on the game’s result, a highly relevant factor in terms of fan interest, or extracting both highlights and lowlights, the latter of which has proven difficult with previous approaches. Lastly, we employ graphical game summaries and their corresponding highlight videos to illustrate the power and flexibility of the proposed framework to produce highlight and lowlight reels in hockey.
2018 Research Papers & Posters
Deep Learning of Player Trajectory Representations for Team Activity Analysis
Authors: Nazanin Mehrasa*, Yatao Zhong*, Frederick Tung, Luke Bornn, Greg Mori
The paper is about identifying player actions and team activities in games using hockey and basketball trajectory data. The framework can be applied to most team sports given the output of any in-game state valuation model, of which there are many.
The research was funded by Sportlogiq and the Natural Sciences and Engineering Research Council of Canada. The patent is owned by Sportlogiq.
2017 Research Papers & Posters
Clustering and Ranking NHL Players Using Location Information and Scoring Impact
Authors: Oliver Schulte, Zeyu Zhao, Mehrsan Javan, Philippe Desaulniers
This paper looks at team sports such as ice hockey and basketball involving complex player interactions. Modeling how players interact with each other presents a great challenge to researchers in the field of sports analysis. The most common source of data available for this type of analysis is player trajectory tracking data, which encode vital information about the motion, action, and intention of players. At an individual level, each player exhibits a characteristic trajectory style that can distinguish him from other players. At a team level, a set of player trajectories forms unique dynamics that differentiate the team from others. We believe both players and teams possess their own particular spatio-temporal patterns hidden in the trajectory data and we propose a generic deep learning model that learns powerful representations from player trajectories. We show the effectiveness of our approach on event recognition and team classification.