⚽ Over/Under Predictions — AI Football Engine

🌐 Live Dashboard → karidasd.github.io/over-under-predictions

---

Stop guessing. Start calculating. This engine wakes up every morning, scans the globe for football fixtures, runs every match through an AI prediction model, and publishes only the high-confidence Over/Under signals — automatically, every single day. No subscriptions. No paywalls. No bullshit.

---

Why This Exists

Every sports betting tool either:

• Sells you "insider tips" for €29/month with zero accountability
• Uses fake backtests and cherry-picked results

This repo does the opposite. Every prediction is logged before the match starts. Every result is graded automatically at midnight. The win rate you see on the dashboard is real and cumulative — updated daily by GitHub Actions.

If it loses, it shows. If it wins, it shows. No hiding.

---

⚙️ How It Works

07:17 UTC — Morning Job                 23:22 UTC — Night Job
───────────────────────                 ──────────────────────
Scan all global fixtures          →     Fetch final scores
Get AI prediction per match       →     Grade each prediction
Filter by xG confidence           →     Update Win Rate
  • xG ≥ 2.7  →  OVER 2.5              Commit results to repo
  • xG ≤ 2.2  →  UNDER 2.5             Dashboard auto-updates
  • xG > 4.5  →  SKIP (amateur)
Publish to live dashboard

The model uses api-football.com's machine learning engine — the same data layer used by professional analytics firms — to calculate expected goals (xG) for every fixture. Only matches where the model has a strong directional signal are published.

---

📂 Project Structure

over-under-predictions/
├── scripts/
│   ├── fetch_predictions.py   # Morning: scans fixtures, generates signals
│   └── fetch_results.py       # Night: grades results, updates history
├── docs/                      # GitHub Pages live dashboard
│   ├── index.html
│   ├── style.css
│   ├── script.js
│   └── data/
│       ├── predictions.json   # Auto-updated every morning
│       └── history.json       # Cumulative track record
├── .github/workflows/
│   ├── morning_predictions.yml
│   └── evening_results.yml
├── engine.py                  # Poisson distribution engine (manual use)
├── bot.py                     # CLI for manual predictions
└── requirements.txt

---

🚀 Run It Yourself

git clone https://github.com/karidasd/over-under-predictions.git
cd over-under-predictions
pip install -r requirements.txt

Get a free API key at api-football.com (100 req/day, no credit card). Rename .env.example → .env and paste your key.

# Predict today's matches
python bot.py predict --league PL

# Grade last week's predictions
python bot.py evaluate --league PL

Supported: PL · PD · SA · BL1 · FL1

---

📈 Sample Output

Match                       xG          Signal       Confidence
──────────────────────────  ──────────  ───────────  ──────────
Arsenal vs Everton          2.45 / 0.81 OVER 2.5     68.2%
Crystal Palace vs Burnley   0.95 / 0.60 UNDER 2.5    74.1%

Evaluation
──────────────────────────  ───────────  Score  Result
Arsenal vs Everton          OVER 2.5     3-1    WON
Crystal Palace vs Burnley   UNDER 2.5    1-1    WON

Win Rate: 100.0% (2W - 0L)

---

⚠️ For educational and research purposes only. Not financial advice. Gamble responsibly.

Built by DARKAIS Data Science.