dkstephanos/3pp-predictions

Exploring alternative ML solutions for generating end-of-season 3-point percentage of NBA players, based on start-of-season shooting statistics.

NBA 3pt% Predictive Modeling

Project Overview

The 3pp-predictions repository is dedicated to exploring advanced machine learning techniques for predicting the end-of-season three-point shooting percentage (3P%) of NBA players. This project leverages early-season shooting statistics to build models that can forecast player performance with higher accuracy. By combining statistical analysis with machine learning, this project aims to provide a nuanced understanding of player performance trends.

Getting Started

Jupyter Notebooks

The repository contains Jupyter notebooks which are the main drivers of the analysis:

• Preprocessing.ipynb

  • Purpose: Extends the base dataset with advanced statistics, visualizes the raw data, and performs LASSO-based feature selection.
  • How to Run: Open the notebook in a Jupyter environment and execute the cells sequentially. Ensure all dependencies are installed.
  • Key Outputs: Extended dataset with new features, visualizations of data distribution and relationships, selected features for model training.

• Prediction.ipynb

  • Purpose: Trains a variety of models to compare performance. Includes evaluation of model effectiveness through relevant statistics and learning/validation/loss curves.
  • How to Run: Similar to the Preprocessing notebook, run the cells in sequence to train models and generate performance metrics.
  • Key Outputs: Trained models, performance metrics (MSE, R² scores, etc.), visual plots depicting model learning and validation.

Full previews of each ran notebook have been persistsed and pushed to the repo for review.

Installation

To set up your environment to run these notebooks, you will need Python installed along with the necessary libraries. You can install the dependencies using:

pip install -r requirements.txt

This command will install all the required Python packages as listed in the requirements.txt file.

Notes and Acknowledgements

This project builds on my previous NBA-related work available on my GitHub profile. Parts of the code and documentation were developed with AI assistance. Statistical definitions and conceptual insights were referenced from NBA Stuffer.

Contributing

Contributions to 3pp-predictions are welcome! Please submit in the form of pull-requests of the main branch.

© 2023 Koi Stephanos. All Rights Reserved.