NBA Sports Betting Using Machine Learning

Overview

This project predicts NBA game winners and totals (over/under) using team stats and sportsbook odds. It pulls team data from 2007-08 through the current season, builds matchup features, and runs trained models to estimate win probabilities and totals outcomes. It also outputs expected value and optional Kelly Criterion stake sizing.

Features

• Moneyline and totals predictions (XGBoost and Neural Net models).
• Expected value calculation and optional Kelly Criterion sizing.
• Odds ingest from supported sportsbooks or manual input.
• Data processing pipeline and model training scripts.
• Flask web app for browsing outputs.

How it works

1. Collect stats and odds: Get_Data pulls daily team stats from NBA endpoints and stores them in SQLite. Get_Odds_Data pulls sportsbook odds and scores from SBR and stores them in a separate SQLite DB.
2. Build game features: Create_Games merges team stats, odds, scores, and days-rest into a training dataset.
3. Train models: XGBoost/NN scripts in src/Train-Models fit moneyline and totals models.
4. Predict today: main.py fetches today’s schedule, builds matchup features, loads trained models, and prints predictions, expected value, and optional Kelly Criterion sizing.

Requirements

• Python 3.11
• Packages: Tensorflow, XGBoost, NumPy, Pandas, Colorama, Tqdm, Requests, Scikit-learn

Install dependencies:

pip3 install -r requirements.txt

Quick start

python3 main.py -xgb -odds=fanduel

Odds will be fetched automatically when -odds is provided. Supported books: fanduel, draftkings, betmgm, pointsbet, caesars, wynn, bet_rivers_ny

If -odds is omitted, the script will prompt for manual odds and totals.

Optional flags:

• -nn run neural network model
• -xgb run XGBoost model
• -A run all models
• -kc show Kelly Criterion bankroll fraction

Flask web app

cd Flask
flask --debug run

Data pipeline and training

# Create/update datasets
cd src/Process-Data
python -m Get_Data
python -m Get_Odds_Data
python -m Create_Games

# Train models
cd ../Train-Models
python -m XGBoost_Model_ML --dataset dataset_2012-26 --trials 100 --splits 5 --calibration sigmoid
python -m XGBoost_Model_UO --dataset dataset_2012-26 --trials 100 --splits 5 --calibration sigmoid
python -m NN_Model_ML
python -m NN_Model_UO
python -m Logistic_Regression_ML --dataset dataset_2012-26_new --trials 50 --splits 5 --calibration sigmoid
python -m Logistic_Regression_UO --dataset dataset_2012-26_new --trials 50 --splits 5 --calibration sigmoid

Neural network notes

• The current NN training scripts are the original versions with hard-coded dataset and model paths.
• They train on dataset_2012-24_new and save into Models/ with timestamped names.
• If you want configurable flags or feature/scaler sidecars, switch back to the newer NN scripts.

Backfilling missing data

Get_Data normally fetches only new dates in the current season. To fill missing dates:

cd src/Process-Data
python -m Get_Data --backfill

To backfill a single season:

cd src/Process-Data
python -m Get_Data --backfill --season 2025-26

Backfilling odds data

Get_Odds_Data normally fetches only new dates in the current season. To fill missing odds dates:

cd src/Process-Data
python -m Get_Odds_Data --backfill

To backfill a single season:

cd src/Process-Data
python -m Get_Odds_Data --backfill --season 2025-26

Contributing

Contributions are welcome. If you change model behavior or data pipelines, add a note in the README and update any related scripts or docs.