Probabilistic pricing for FIFA World Cup 2026. M★ is a bivariate Poisson model with Dixon-Coles correction, calibrated on international match data. Model-implied probabilities come from 10,000 Monte Carlo simulations per nightly snapshot; a market-comparison layer de-vigs bookmaker odds and reports signed model-vs-market divergence in the divergence terminal. It covers knockout fixtures live: model probabilities conditioned on settled results are compared against de-vigged market odds, ungraded. Only the 72 pre-registered group-stage forecasts are ever graded. When the odds are older than 30 hours the layer reports itself stale rather than showing stale rows. The “45% problem” refers to a systematic divergence documented in Phase 1: market-implied championship probabilities for mid-tier contenders cluster near 45% of their model-implied values, suggesting a persistent structural mispricing that motivated this research programme.
Pre-registered at osf.io/spmkg · tag v1.0.0-mstar-lock · Phase: completed · 0 matches remaining

A mathematical representation of 10,000 Monte Carlo samples from the M★ posterior, projected onto the FIFA World Cup trophy.
Every number on this page is traceable to a snapshot and a code SHA in the Divergence Terminal.
Pick four semifinalists. The model has run 10,000 simulated tournaments. See where your scenario lands.
FROZEN PRE-TOURNAMENT FORECAST. Computed before the opening match and not updated with results. The live conditional view lives on the Bracket page.
FROZEN PRE-TOURNAMENT FORECAST. Computed before the opening match and not updated with results. The live conditional view lives on the Bracket page.
Each match shows the posterior probability that the favoured side advances. Click a team to open its progression page. The chain · R16 through final: is the sequence of modal outcomes; its joint likelihood is much lower than any leg alone.
Markets where M★ disagrees most with de-vigged book consensus.
No real bookmaker lines are ingested yet, so no divergence rows are published. Real de-vigged divergence appears once the odds producer runs.
Largest 7-day shifts in Δ rank. Reported descriptively: pre-registered §7.2.
| Team | P(champion) | Δ rank · 7d |
|---|---|---|
Brazil | 6.3% | +3 |
Netherlands | 4.8% | +2 |
| Team | P(champion) | Δ rank · 7d |
|---|---|---|
Germany | 3.8% | −2 |
Spain | 18.2% | −2 |