Set salaries for minor league hockey players.
Students: Joshua Dixon, Kayko Ohkawa, Spencer Segal
The data analytics unit of the Pittsburgh Penguins professional hockey team enlisted the capstone team to develop an empirical model to set appropriate salaries for minor league hockey players. While the Penguins had a model for setting salaries of players in the major league, they did not have a model for salaries of players in lower level minor leagues, and to their knowledge, neither does any other professional hockey team organization. The project involved merging salary data for minor-league players with data from other sources on player characteristics and player game statistics. Merging the datasets involved a variety of techniques, including fuzzy-matching. After initial data visualization, the team developed a series of prediction models using a range of different techniques, from linear modeling to random forest machine learning. The resulting models can help answer a key business question, which is how to set salary amounts based on player characteristics and market conditions, avoiding overpaying but also avoiding losing good players. The team developed an interactive version of the model that allows users to enter player names and receive the predicted salary information. PDF Presentation.
Client reviews by Sam Ventura, PhD., Director of Hockey Operations and Hockey Research, Pittsburgh Penguins
- How did the MQE students meet the demands of your project? They had the preparation (coursework, knowledge, skills, etc) necessary to complete the project. They worked together as a team to successfully navigate a complicated data science problem (from data acquisition, exploration, visualization to modeling and communication of results).
- What was the most satisfying professional experience that you had with the MQE students? I was most impressed by their ability to meet the challenge that I gave them with one week to go before their final presentation. I asked them to turn their entire data pipeline (including modeling, results, and then out-of-sample predictions) into an interactive program where I could input new players and get estimated salaries. I didn't think they would be able to pull it off because of how difficult of a task it was, but they did. Best of all, the rest of the work that they had to complete did not suffer at all, culminating in a fantastic final presentation.
- What is unique about the MQE program and its training of future innovators in economics and data science? The MQE program specializes in the combination of data science and economics. Unlike other programs, the MQE program not only prepares students with the data science skills necessary to enter the modern data science workforce, but it provides a rigorous background in economics and related topics, setting MQE graduates apart from graduates of more singular-focused programs.