Check out some of my other work in my 'AdditionalProjects' repository: https://github.com/DanMccue/AdditionalProjects

NHL Player Salary Analysis Project

Overview:

• A machine learning analysis of NHL player salaries using performance metrics to predict compensation and identify market inefficiencies, using XGBoost and Lasso regression models.

Dataset:

• 789 NHL players from 2016-17 season

• 154 performance features including:

  • Traditional stats (Goals, Assists, Points)

  • Advanced metrics (Fenwick, Corsi)

  • Physical attributes (Height, Weight)

  • Career statistics (Experience, Draft position)

  • Salary range: $575K to $14M

Methodology:

Data Preprocessing

• Age calculation from birth dates

• Experience derivation from draft years

• Position standardization (Defense/Center/Wing)

• Draft position normalization

• Advanced metric computation

• Handling of undrafted players

Model Development

• Implemented XGBoost and Lasso regression models

• Cross-validation with 5 folds

• Hyperparameter optimization

• Feature importance analysis

Results:

Model Performance

• Lasso Regression achieved 47.0% accuracy (R² = 0.470 ± 0.030) with RMSE of $1.69M

• XGBoost achieved 44.5% accuracy (R² = 0.445 ± 0.058) with RMSE of $1.73M

Key Predictors

• Flow Statistics

  • Fenwick For (FF): Strongest predictor

  • Corsi For (CF): Third most important

  • Individual Shots (iSF)

• Scoring Metrics

  • Goals Scored (GS): Second strongest predictor

  • Expected Goals For (xGF)

  • Points (PTS)

Salary Analysis

• Most efficient team: Carolina Hurricanes underpaying by $0.84M per player average

• Least efficient team: Buffalo Sabres overpaying by $0.55M average

Notable Contract Discrepancies (predicted salary in millions)

• Highest Overpaid: Stamkos ($7.83M), Kopitar ($7.52M), Toews ($7.21M)

• Most Underpaid: Panarin ($4.37M), Slavin ($4.34M), Gostisbehere ($4.27M)

Market Inefficiencies

• Entry Level Contracts significantly undervalue young talent

• Veteran contracts often exceed statistical performance value

• Teams with efficient salary structures tend to maximize performance metrics relative to payroll

Limitations:

• Single season analysis

• Actual salary vs cap hit consideration

• Unquantifiable factors (leadership, marketability, reputation)

• Entry-level contract restrictions

Author: Daniel Mccue

• Last Updated: January 2025