four_factors

"All models are wrong, but some are useful." - George E.P. Box

Introduction

This research intends to:

1. Introduce Dean Oliver's "Four Factors of Basketball Success in the NBA" and test if the four factors are statistically significant predictors of success in the NBA
2. Use the four factors to predict a team's number of wins and average margin of victory in the NBA using historical data and various machine learning models
3. Compare the observed weightings of the different factors in our models to those proposed by Dean Oliver and Ed Küpfer.

What's In This Repository?

Report

• thesis_draft.pdf
• thesis_draft.zip
  • main.tex
  • images

Notebooks

• Modeling Oliver’s Four Factors.pdf
  
• R_four_factors.Rmd
  
• basketball_reference_scraping.ipynb
  
• historical_changes_visualizations.ipynb
  
• modeling.ipynb
  

Datasets

• per_100_posessions_historical.xlsx
• four_factors_20xx_to_20xx.xlsx
• four_factors_all_seasons.xlsx

Figures

Models Used

Multiple Linear Regression
Random Forest
Gradient Boosting
Neural Network (future work)

R Libraries Used

Analysis

sjstats

Data Processing

dplyr
readxl
tidyr

Data Visualization

corrplot
formattable
GGally
ggplot2
kableExtra
knitr
plotly
scales
stargazer
viridis

Machine Learning

car
caret
gbm
randomforest

Python Libraries Used

Data Processing

NumPy
Pandas

Data Visualization

matplotlib
seaborn

Machine Learning

Scikit-Learn
Statsmodels

Additional Resources

An Introduction to Statistical Learning: with Applications in R

• Resource for understanding statistical learning models

The Elements of Statistical Learning

• Resource for understanding statistical learning models in greater depth

Basketball Reference

• Resource for obtaining NBA data