Tarmily Wen and Andrew Petrosky
This is a machine learning predictor for NBA games.
We built a scraper to get data from Basketball Reference.
We tested with a variety of different models, ranging from a simple baseline predictor which just predicts each game based off of the current win percentage of each team to models such as Gaussian Naive Bayes and SVM.
Best test accuracy was 74.3% from SVM so far, compared to 69% from the baseline predictor.