Analysis of hockey forward line Corsi: should the focus be on forward pairs?
Faculty Advisor
Date
2024
Keywords
K-Means, Random Forest, Support vector machine, sports data, NHL
Abstract (summary)
Professional ice hockey is a popular sport in North America, with multiple previous analyses providing insights into teams. Most research has been done on analyzing pairs of players on the same team that work well together. The focus of this study was to analyze if trios on a forward line perform well together, as there has not been enough research in this field. Our goal was to determine if the third player changes the performance of a duo and identify key factors that explain this change. We have analyzed more than 14 years worth of data. This data started with more than 100 dimensions; from those 100, 35 dimensions were chosen for analysis. To reach our conclusion, we used three methods: K-Means, Random Forest, and Support vector machines. Single variate random forest was used to analyze which variables affected the Corsi Percentage. The results from K-Mean clustering, combined with the results from Single Variate Random Forest, were used to see if the substitution of a third player on a line of three makes a difference in the overall performance of the line. The Support Vector Machine algorithm was used to reinforce the cluster numbers obtained from K-means clustering. Our study found that adding a third player will have a positive effect when the third player consistently plays with the other two players and the three players participate more effectively in defence. These findings could help teams plan how they form their player lines when they want to achieve good game results.
Publication Information
Brownlee, S., Khan, A., Vanderzyl, B., & El-Hajj, M. (2024). Analysis of hockey forward line Corsi: Should the focus be on forward pairs? In 2024 IEEE 14th Annual Computing and Communication Workshop and Conference (CCWC), Las Vegas, NV, USA, 2024, pp. 0242-0248. https://doi.org/10.1109/CCWC60891.2024.10427925
Notes
Item Type
Presentation
Language
Rights
All Rights Reserved