How Futbol Modelo Works

This isn't a static demo — it's a real, running system.

Every morning, an automated pipeline:

1. 1.Pulls fresh match data across four leagues (MLS, Premier League, Serie A, La Liga).
2. 2.Generates predictions using two model families — a Dixon-Coles statistical model for match outcomes, and gradient-boosted models (LightGBM) for player and team props (corners, shots on target, goalscorers, saves). Every model had to prove it beats a naive baseline before being allowed to ship.
3. 3.Locks each prediction permanently the moment it's made, in an append-only ledger nothing can edit or delete — even the system itself. That's what makes the track record real: there's no way to quietly fix a wrong call after the fact.
4. 4.Grades every result automatically once the match ends — hits and misses alike, published either way.

It all runs unattended on a self-managed Kubernetes cluster — no manual intervention required to keep it going.

Alongside the pipeline, Petey answers real questions about the data — how a market has performed, how a team has been playing lately. Petey never writes its own database queries or sees raw rows: it only ever phrases a pre-computed, validated answer as a sentence, and it says so plainly when a sample is too small to trust.

Futbol Modelo is an analytics and calibration project, not a gambling product. It doesn't place bets, take odds, or offer betting advice — every prediction exists to be graded against reality, nothing more.