Jupyter notebook that outlines a simple predictive model for NBA performance. Predicts the peak "WinShares" by the current draft prospects based on numerous features such as college stats, projected draft pick, physical profile and age. I try out multiple models and pick the best performing one for the data from my judgement.
Uses linear regression, a neural network multi-layer perceptron regressor, ridge regression, lasso regression and a linear support vector regression.
All models were created with scikit-learn.
Database was built from scratch from RealGm, Basketball Reference and NBA.com.
*Since database was built from scratch I don't have as much data points as I'd like so model results experience pretty significant variance.
*WinShares is a metric that has flaws so not a perfect measure for predicting NBA success.