100,000 Monte Carlo simulations based on current team ratings
How our model's World Cup 2026 championship probabilities compare against sportsbook consensus and prediction markets. We use disagreements as signal, not noise.
| # | Team | Our Model | Consensus▼ | Diff |
|---|---|---|---|---|
| 1 | Spain | 100.0% | 65.2% | +34.8* |
| 2 | Argentina | — | 34.8% | -34.8* |
2 teams
Teams where our model diverges meaningfully from market consensus. These are interesting signals, not necessarily errors.
Model: Grid-Optimized Composite (60% Bradley-Terry, 30% Elo, 10% Roster) | 50,000 Monte Carlo iterations | Simulated August 15, 2026
Sportsbook odds: US books, best-available across 1 book(s) (updated 2026-07-19) | Polymarket: $4.33B traded (updated 2026-08-15)
Consensus = average of sportsbook implied probability and Polymarket probability, normalized. Diff = model minus consensus (positive = our model is more bullish on that team). * = difference is statistically significant beyond Monte Carlo sampling noise (95% CI).
Note: Our composite weights and prediction sensitivity were calibrated against these same market odds. The metrics above measure in-sample calibration fit, not independent validation. See backtesting results for out-of-sample predictive accuracy.
