AN AGENT-BASED SIMULATION FRAMEWORK FOR ANALYZING TRAFFIC EFFECTS IN ICE HOCKEY

Sammanfattning

Analytics in hockey is a big part of the game today but the proprietary nature of tracking data could halt the development. Traffic - when offensive players position themselves in front of the opposing goaltender to obstruct vision and redirect shots - is a widely used strategy but remains largely understudied analytically. This study explores the use of agent-based simulation as an alternative to the proprietary tracking data, using NetLogo to generate synthetic shot data which is analyzed using logistic regression. The simulation is calibrated against empirical results from existing literature through hyperparameter optimization across 7800 parameter combinations. Results show that the simulation successfully replicates real-world patterns for mid and long range shots. A novel traffic position metric is introduced to investigate how the positioning of players creating traffic affects shot outcomes. Logistic regression with an interaction term between traffic position and shooter zone reveals that traffic positioned close to the goaltender significantly increases goal probability for mid and long range shots, with the effect strengthening as shooter distance increases. These findings suggest that traffic positioning is a meaningful factor in goal scoring and that agent-based simulation can capture certain patterns from real hockey, but more sophisticated mechanics are needed before it can be considered a reliable substitute for real tracking data.