NBA Game Predictor

As a Toronto Raptors fan I really wanted to somehow give myself realistic expectations for the 2018 playoffs after the dissapointment of the series against the Cavs last year. This desire to give me some sort of reasonable expectations combined with my intersest in data and machine learning spawned the idea of making an NBA Game Predictor.

Project Screenshots

ELO values for home and visiting teams vs. the outcome of the game in the 2016-2017 NBA season

Change of logloss vs. the number of data points(predictions made using a 2 layer 46 node neural network on the test set)

Project Status

Completed:

• write scraper to collect raw datasets from www.basketball-reference.com
• write script to clean raw data and get it ready to for feature addition
• write script to add features to cleaned data
• implement a multi layer neural network and start predicting games
• write a script to calculate evaluation metrics for ML algos
• clean up/modularize data_preparer.py
• implement Nate Silvers ELO algorithm for NBA teams

In Progress:

• create seperate script for all data viz methods

To Do:

• write and apply mean normalization func to all features
• implement multivariate logistic regression to predict games
• implement decision tree to predict games
• create GUI to easily see predictions for a specific day

Installation and Setup Instructions

Still have to update this :)

Reflections

I'll update this when I'm done :)

Technologies Used