NBA-Machine-Learning-Sports-Betting

Why we included this project

This repository is a working reference for anyone assembling a sports-betting pipeline. It pulls daily NBA team stats back to the 2007-08 season, merges them with sportsbook odds and scores into matchup features, and trains XGBoost and neural network models to predict moneyline winners and over/under totals. What bettors will appreciate most is the expected value calculation and optional Kelly Criterion stake sizing, which turns raw predictions into a concrete decision about whether a wager is worth placing. The code is split into separate modules for data collection, feature building, training, and prediction, so you can trace how historical game data becomes a betting signal. A small Flask app lets you browse the outputs instead of working only from the command line.