One AI model saw it coming.
Across 63 scored group stage matches, Claude Opus 4.8 racked up 62 points with 12 exact scorelines and 26 correct results — finishing 15 points clear of its nearest competitor in the benchmark. It was the only model to predict all 104 matches without a single gap or hallucinated team. What you're seeing isn't luck: it's a methodology that weighted recent form, squad fatigue, altitude, and travel friction into every single prediction. It finished 13 points ahead of the runner-up.
Claude Opus 4.8 answered these questions before making a single prediction. This is how it thought about the tournament.
Ensemble of Elo, Dixon-Coles, and historical WC; recent tournaments weighted 50%, WC history 35%, current squad depth 15%.
USA +18% (massive home crowd factor), Mexico +22% (Azteca effect), Canada +10%.
Full exogenous model: squad fatigue, altitude (Mexico City), travel time, and injury probability based on club season minutes.
Starting from the Round of 32, Claude Opus 4.8 predicts each knockout round after seeing the actual qualifiers — the same AI model, the same methodology, real-world knowledge applied round by round. The leaderboard freezes the a priori race between all 10 models; this bracket is the live experiment with the champion.