Odds Arbitrage Scanner

Real-time arbitrage scanner across 30+ international sportsbooks. Aggregates decimal odds from the-odds-api.com, computes book sums across best-per-outcome cross-bookmaker lines, and flags cases where the combined implied probability drops below 1 — mathematical sure-bets.

Also detects value bets by using Pinnacle (the sharpest bookmaker in the industry) as a "true probability" anchor and finding other books offering longer odds for the same outcome.

Exposes results as both a Rich terminal CLI and a FastAPI REST service.

What counts as arbitrage

For any event with outcomes, a bookmaker's odds imply a probability of 1 / odds. Summing these across all outcomes gives the book sum. Bookmakers always price with book_sum > 1 (their margin).

But different books disagree on outcome probabilities. If you take the highest odds across different books for each outcome, the combined book sum sometimes drops below 1. Bet proportionally on each outcome and you're guaranteed a profit no matter who wins.

guaranteed_profit % = (1 - book_sum) × 100
stake_i            = total_stake × (1 / odds_i) / book_sum

Typical real-world arbs: 0.5%–3% edges. Rarely above 5% outside of boosted promos or pricing errors. Windows close quickly (often in minutes), so you need to be fast.

Example run

Scanned 342 matches across 21 sports  ·  10 arbitrage opportunities found

#1  Burnley vs Nottingham Forest  (Premier League, 3-way)
  Profit: +$4.12 on $100 stake  =  +4.12%   (book sum: 0.9604)

   Bet                │ Bookmaker     │ Odds   │ Stake   │ Return
   ───────────────────┼───────────────┼────────┼─────────┼────────
   Burnley            │ Betfair       │ 7.800  │ $12.94  │ $100.93
   Nottingham Forest  │ GTbets        │ 1.550  │ $65.12  │ $100.93
   Draw               │ Winamax (DE)  │ 4.600  │ $21.94  │ $100.93

   → Place all three bets, whichever result occurs you collect $100.93.
   → Net profit: +$4.12 (after subtracting $96.81 combined stakes? no - wait,
     the example uses $100 total distributed across legs)

Actual tested run (April 18, 2026): found 10 profitable arbitrages across EPL, La Liga, and Bundesliga — combined +$12.47 profit on $1000 of stake.

Architecture

odds-arb-scanner/
├── src/
│   ├── odds_client.py   # the-odds-api client + in-memory TTL cache
│   ├── arbitrage.py     # Math engine (2-way/3-way arb + value-bet detector)
│   ├── display.py       # Rich terminal tables
│   └── api.py           # FastAPI app (/arbs, /value, /sports, /health)
├── docs/
│   ├── screenshot.png
│   └── make_screenshot.py
├── main.py              # CLI entry (scan | list | serve)
├── requirements.txt
└── .env.example

Tech Stack

• Python 3.10+ — asyncio, dataclasses, type hints
• httpx — async HTTP with timeouts
• Rich — terminal UI
• FastAPI + uvicorn — REST API
• python-dotenv — env config

Installation

git clone https://github.com/cengizmandros/odds-arb-scanner.git
cd odds-arb-scanner
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env
# add your the-odds-api.com key

Free API tier: 500 requests/month at https://the-odds-api.com/

Usage

CLI scan

# Scan all default sports for arbs
python main.py scan

# Custom stake + minimum profit threshold
python main.py scan --stake 500 --min-profit-pct 1.0

# Specific sports only (saves API quota)
python main.py scan --sports soccer_epl,soccer_spain_la_liga

# Include value-bet detection (requires Pinnacle coverage)
python main.py scan --value --min-edge-pct 2.0

# List available sports
python main.py list

REST API

python main.py serve --port 8000

Endpoints:

• GET / — landing page with endpoint docs
• GET /arbs?stake=100&min_profit_pct=0.5 — arbitrage opportunities
• GET /value?stake=100&min_edge_pct=2.0 — value bets vs Pinnacle
• GET /sports — available sports
• GET /docs — auto-generated OpenAPI documentation
• GET /redoc — alternate API docs

Example API response

$ curl http://localhost:8000/arbs?stake=100&min_profit_pct=1.0

{
  "stake": 100,
  "arbs_found": 3,
  "arbs": [
    {
      "home_team": "Burnley",
      "away_team": "Nottingham Forest",
      "market_type": "3-way (1X2)",
      "profit_pct": 4.12,
      "guaranteed_profit": 4.12,
      "book_sum": 0.9604,
      "legs": [
        {"outcome": "Burnley", "bookmaker": "Betfair", "odds": 7.8, "stake": 12.94},
        {"outcome": "Nottingham Forest", "bookmaker": "GTbets", "odds": 1.55, "stake": 65.12},
        {"outcome": "Draw", "bookmaker": "Winamax (DE)", "odds": 4.6, "stake": 21.94}
      ]
    },
    ...
  ]
}

Practical considerations

1. Bookmaker access: the-odds-api aggregates 30+ books, but you'll only have accounts with a subset. Filter the output to bookmakers you can actually bet at.

2. Odds drift: by the time you see an arb, prices may have moved. Execute quickly or set high thresholds to leave headroom.

3. Account limits: bookmakers hate arb bettors and will often limit or ban accounts that win consistently using cross-book strategies. Betting exchanges (Betfair, Matchbook) don't care because you're playing against other users.

4. Starting bankroll: arbs profit in the 0.5–3% range. You need decent capital to generate meaningful absolute returns — $100 × 1% = $1 per arb.

5. Cache policy: the client caches odds for 180s by default to conserve free-tier API quota. Tune via CACHE_TTL env var.

Roadmap

• Historical odds tracking (SQLite) — detect "closing line value" over time
• WebSocket subscription for real-time updates (where supported)
• Telegram bot integration for instant alerts
• Auto-filtering by accessible bookmakers (regional preferences)
• Kelly criterion sizing for value bets
• Historical backtest engine

License

MIT