Real model. Public grades. No cherry-picking.

What Edge Engine is

Edge Engine is a small, solo-built prediction service for U.S. sports. It's not a tipster newsletter, a Discord "lock" service, or a black-box AI marketing product. It's a live statistical model that pulls sportsbook lines, computes what it thinks each game's real probability is, and stores every prediction with a timestamp before the game starts.

Then every game is automatically graded against the sportsbook result. Every prediction is on the record, whether it won or lost.

The model, in plain English

For each upcoming game, Edge Engine combines:

• Team strength ratings — per-sport rating systems for the in-season leagues, plus surface-aware ELO for ATP and WTA tennis
• Sportsbook line movement — how odds shifted from open to close is often more informative than the price itself
• Injury and lineup data — starting pitchers, key injuries, days of rest
• Historical calibration — a per-sport correction (Platt / isotonic) so a stated "60% win probability" actually resolves to ~60% in real outcomes

These feed a probability estimate. If the model's estimate is meaningfully higher than the sportsbook's implied probability, that's a pick.

Why the grades are trustworthy

The grading is what separates Edge Engine from most services in this space. Every prediction is written to the database before the game starts, with a locked model version and timestamp. When the game ends, an automated grader looks up the result from The Odds API and marks the prediction correct or not.

No prediction gets edited after the fact. No losing picks get quietly deleted. The accuracy page shows the aggregate; the recent picks page shows every recent game with a green ✓ or gray ✗ next to it. Nothing is hidden.

Who's behind this

One person. Solo engineering, solo modeling. No team, no VC, no marketing agency. Every line of code, every model retrain, every bug fix. If you email support you're emailing the person who built the thing.

The site has been in continuous operation for over six months. Model retraining is automated (weekly per-sport Platt calibration; periodic feature and ELO refreshes). Every design choice is documented so future improvements don't undo past ones.

What Edge Engine is not

• A guaranteed win. A ~60% win rate means you'll lose ~40% of the time. That's normal. Bad weeks happen.
• A get-rich-quick service. Compounding at 5-15% edge is slow. Anyone promising 30%+ ROI monthly is selling something else.
• A sportsbook. Edge Engine sells predictions. Placing the actual bet is on you, with a licensed sportsbook.
• Legal advice on gambling. Check your local laws before betting.

Ready to see the picks?

The accuracy page and recent picks are public. The full daily picks, model probabilities, confidence intervals, and reasoning are behind the paywall.

Start with a 7-day free trial — no credit card required. If you don't add one by day 7, access simply stops. No auto-charge, no phone call to cancel.