You Miss 100% of the Attendance You Don’t Optimize: A Data-Driven Approach to NHL Scheduling

Abstract

This thesis aims to optimize the National Hockey League's (NHL) game schedule through the maximization of attendance, while adhering to the strict constraints that the NHL schedule must follow, such as arena availability, match-up constraints, and fairness. By analyzing how various factors, such as team statistics from previous seasons, game date, and location, influence game attendance, this thesis aims to optimize the schedule for game attendance, while maintaining fairness, practicality, and feasibility. Through detailed analysis of attendance at NHL games during past seasons, this thesis first identifies what features impact game attendance and chooses an attendance prediction model. This attendance prediction model is subsequently used to predict game attendance for the 2024-25 season. These predictions are then used to construct an attendance optimization model. Although the results of the attendance optimization problem showed only marginal improvements over the current schedule, they highlight the possibility of implementing a more data-driven approach to sports scheduling.