A deep dive into NBA stats and a model to predict the outcome of games.
Check out this medium article (Currently being published) I wrote diving into some of the analyses and describing the model.
processed-data/
contains all the data used in exploration. It includes data such as games, salary, player stats, etc. Many of the data spans the past 5 seasons 15-16 to 19-20. It is scraped from different sources such as nba.com and basketball-reference.com. full-data/
contains processed train/test data for the model.
The data is explored in both SQL and Tableau. SQLite3 for Python is used, as a light-weight SQL engine. Make sure you also have pandas installed. You can try Tableau for free or get it with a student email.
fit-model.ipynb
contains the code for the model which uses XGBoost in Python. It also explores the outcomes of the trained model.