Ratings. Every club's strength comes from its public Elo rating (clubelo.com), refreshed daily. Home advantage is worth 65 Elo points.
Match model. An Elo gap becomes goal expectations through a bivariate Poisson β two correlated goal processes, the same mathematics that ran the WCPA World Cup model.
The season, simulated. Thousands of full 380-match seasons are played out, conditioned on every real result so far. A summer rating is not a season, so each simulated season draws every club's "true self" with honest noise (Ο = 55 Elo) β no fake certainty in July.
Frozen calls. Before each matchweek kicks off, the model's 1X2 call and most-likely score are frozen β append-only, timestamped, never rewritten. Results grade them automatically. The record, hits and misses, lives on this page forever.
Lineage. This engine called 71/104 (68.3%) at World Cup 2026 β including Spain in the final at 51.3% of a coin flip. PREMPA is the same honest machine, given a league season to live in.