Undergraduate Honors Thesis


Using Machine Learning and Traditional Statistical Approaches to Predict Hockey Goals in the National Hockey League Public Deposited

Downloadable Content

Download PDF
  • This thesis investigates the accuracy of a machine learning approach and a traditional statistical approach in predicting goals in the National Hockey League (NHL) and identifies the factors that are significant in goal prediction. The topic of sports analytics in the NHL is under-researched, evidenced by a gap in the available literature on using analytics in the NHL. The limited studies on the topic share the consensus that shot-based metrics and goal-based metrics regarding NHL-level hockey data can be built into both traditional statistical models and machine learning models for prediction using logistic regression. The data used in this thesis consists of historical NHL shot data sourced from Moneypuck.com for the Colorado Avalanche over a three-season period from 2016 through 2018. Using Python and logistic regression, a machine learning prediction approach was built with scikit-learn and a traditional statistical prediction approach was built with statsmodels. After an in-depth inspection, the analysis shows that the prediction had an overall test accuracy of 82%, with 45% of goals scored predicted. The findings of this study identified the most significant variables in the prediction model and increasing goals include home team goals, shot rebound, home skaters on ice, away penalty length, home penalty length, shooter time on ice, shot on goal, the last event being a give, the last event being a take, the last event category being a penalty, and the player position being defenseman. The findings of this study can be widely used in the NHL to create profitable outcomes and investments for the league, increase the appeal for future partners and stakeholders, improve conditions for teams and players, and ensure the vitality of the NHL.

Date Awarded
  • 2022-04-08
Academic Affiliation
Committee Member
Granting Institution
Last Modified
  • 2022-04-18
Resource Type
Rights Statement


In Collection: