Project 2: Linear Regression Model for Chris Paul Player Statistics
The project aimed to build a model to help Chris Paul, a professional basketball player, identify the statistics that most influence his offensive ability, particularly points scored per game.
Over 1,000 rows of game statistics were scraped from Chris Paul’s page on Basketball Reference to build a linear regression model.
The model was designed to provide interpretable and actionable insights that Chris Paul could incorporate into his playing strategy.
The data collection involved scraping per game statistics for Paul’s entire career, totaling 1155 regular season games.
A pairplot graph was built to visualize the data and identify any significant collinearities, after which redundant features were removed.
Three different regression models were built: Linear Regression, Polynomial Regression, and Lasso Regression. The Linear Regression model was chosen for its simplicity and interpretability.
The data consisted of Paul’s in-game stats (points, assists, steals, etc.) along with some categorical data such as home vs away, opponent, and date.
The Linear Regression model revealed that steals and defensive rebounds were not significant predictors of points per game as their p-values were greater than 0.6.
Field goal attempts and free-throw attempts were found to be the biggest indicators of points scored, with coefficients of 1.11 and 0.8 respectively, followed by three-point attempts with a coefficient of 0.39.
Interestingly, offensive rebounds were found to be the largest deterrent of points scored, with a coefficient of -0.914.
