The record.
Fantasy squad managed across 8 gameweeks. Spain backed from the group stage. 104 match predictions — all committed before kick-off, all on the record.
Who does the model back?
Pick any two nations. The Dixon-Coles model scores them.
FPL tools, powered
by the same model.
The same Dixon-Coles engine that predicted and graded all 104 WC2026 matches turns to FPL next. Every squad committed before the deadline, every score graded in public — including the bad ones.
The first squad is committed about ten hours before the Gameweek 1 deadline. Its fingerprint appears here the moment it's published.
The model's gameweek scores appear here as they are graded — every squad committed before the deadline, every score published afterwards, including the bad ones.
The model plays by your rules — one squad, real budget, real transfers, hits paid. See how you did against it.
Every match. Every gameweek. Every call.
Built on a proven forecasting engine.
We tested three models on GWs 29–34 of 2025/26 — weeks no model trained on. The Dixon-Coles projection won. The ML model performed no better than a naive baseline on an honest out-of-sample test, so we didn't ship it. The same approach now runs every WC 2026 fixture and the 2026/27 FPL season.
Built on football analytics,
tuned to WC 2026.
TheModelSays uses a Dixon-Coles Poisson model, the same statistical framework used in academic football research and professional betting markets. Here is how it works.
Dixon-Coles Poisson model
At the core is a Dixon-Coles model — the statistical standard for football prediction. It estimates each team's attack strength and defensive weakness from a weighted blend of WC 2018, WC 2022, and 2,000+ recent international results. A low-score tau correction improves accuracy on 0-0, 1-0, 0-1, and 1-1 scorelines, which a naïve Poisson model under-predicts.
ELO ratings + form adjustments
Alongside the DC model, ELO ratings are computed from the same match history, capturing momentum across time. Form adjustments then compare each team's actual goals scored and conceded in their last 8 matches against the DC baseline — if a team is consistently outperforming expectations, the model boosts their attack rating accordingly.
50,000 Monte Carlo simulations
The tournament simulation draws 50,000 independent runs of the full World Cup — all 72 group-stage games, the Round of 32, Quarter-Finals, Semi-Finals, and Final. Group-stage results that have already been played are pinned to the actual result. The win probability shown for each team is simply how often they lift the trophy across those 50,000 paths.
Walk-forward retraining — tested and dropped
We retrained after each matchday, weighting WC 2026 results at 4× over historical data. Graded across all 104 matches, it made things worse: the frozen pre-tournament model called 66.3% of outcomes correctly against the retrained version's 59.6%. The gap was widest in the knockouts — 68.8% versus 53.1%. Walk-forward retraining has been dropped for the FPL season.
TheModelSays is an independent, free-to-use football prediction tool. The WC2026 model is complete — 104 matches, all on the record. Next up: FPL 2026/27, launching with the new Premier League season. Not affiliated with FIFA or the Premier League. All predictions are generated by a statistical model for informational and entertainment purposes only. Follow updates on TikTok and YouTube.
TheModelSays104 WC2026 match predictions — all on the record. The same model turns to FPL 2026/27 next.