Sports Betting Dashboard — NHL & NBA

Disclaimer: This project is for informational and educational purposes only.
It is not investment or betting advice.

Overview

This repository powers a public-facing day-ahead prediction dashboard for NHL and NBA markets — including:

• Moneyline
• Totals (Over/Under)
• Asian handicaps

Motivation

This project started as a curiosity: can a well-engineered machine learning system match or outperform market closing lines using only on-court performance data?
It’s primarily a learning and research playground focused on time-series modeling, feature engineering, and execution strategy optimization, wrapped in a transparent, versioned prediction system that anyone can evaluate over time.

Technical Stack

• Backend / Modeling: Python (pandas/polars, scikit-learn, tensorflow, kedro, dash)
• Data Storage: PostgreSQL / parquet artifacts
• CI/CD: GitHub Actions, ArgoCD, kubernetes
• Frontend:
  • (on-prem) Dashboard with up-to-date predictions (dash)
  • (public) Static GitHub Pages (HTML + JS)

Majority of the solution is hidden in private repositories at this time. If the models are not good enough to beat the markets continuessly I will release all data pipelines along with model repository.

Release cadence

• Predictions updated daily (day-ahead)
• Results added the following day

Data conventions

One JSON object per prediction/result (filename: {YYYY-MM-DD}_{HOME}_{AWAY}.json), e.g. 2024-10-30_UTA_LAL.json

Predictions : docs/data/{league}/predictions, under the key "preds" are probability predictions for various events.

{
  "match_id": "123",
  "start_time_utc": "2024-10-30T23:30:00Z",
  "home_team": "Utah Jazz",
  "away_team": "Los Angeles Lakers",
  "preds": {
    "rt": {"home": 0.5231, "draw": 0.0268, "away": 0.4502},
    "ft": {"home": 0.5375, "away": 0.4625},
    "mu": {"home": 110.8277, "away": 109.4692},
    "std": {"home": 10.9922, "away": 15.3127},
    "hcp": [{"hcp": -0.5, "prob": 0.5231}, {"hcp": 0.5, "prob": 0.5498}],
    "ou": [{"points": 220.5, "prob": 0.4962}, {"points": 221.5, "prob": 0.4774}]
  }
}

Results: docs/data/{league}/results, contains only match id and final scoreline.

{
  "match_id": "123",
  "home_team": 2, 
  "away_team": 1
}