Who wins?
Spain v Argentina, July 19 at MetLife. Spain shut out France; Argentina beat England 2–1 from behind. At kickoff I triangulated Opta, Silver's model and the market: Opta had Spain 59.5 to 40.5 and split the 90 minutes 45 / 29 / 26, so nearly a third of its runs went to extra time. The numbers favored Spain 55–60 to 40–45; my heart, Argentina. Spain won 1–0 in extra time (Torres 106′), exactly the tail that split had flagged.
Snapshot as of July 21, 2026
EloRating turning strength gap into win probability.Monte CarloEstimating outcomes from thousands of random simulations.overroundBookmaker's margin; implied odds top 100%.
Quarterfinals
🇦🇷 Argentina 3–1 SUI
🏴 England 2–1 NOR
🇫🇷 France 2–0 MAR
🇪🇸 Spain 2–1 BEL
Semifinals
🇫🇷 France 0–2 Spain 🇪🇸
- ⚽ 22' Mikel Oyarzabal
- ⚽ 58' Pedro Porro
🏴 England 1–2 Argentina 🇦🇷
- ⚽ 55' Anthony Gordon
- ⚽ 85' Enzo Fernández
- ⚽ 90'+2' Lautaro Martínez
Final
🇪🇸 Spain 🏆
The models, side by side
| Team | Opta title | Opta reach final | Bookies | Author range Speculative |
|---|---|---|---|---|
| 🇪🇸 Spain | 59.5% | 100.0% | -170 (63%) | 55%–60% |
| 🇦🇷 Argentina | 40.5% | 100.0% | +135 (43%) | 40%–45% |
| 🏴 England | 0.0% | 0.0% | — (0%) | 0%–0% |
| 🇫🇷 France | 0.0% | 0.0% | — (0%) | 0%–0% |
Nate Silver's model is paywalled, so it isn't in the table — more in the fine print.
Sources:Opta / The Analyst — 2026 supercomputer predictions ↗ · TNT Sports — Opta's supercomputer refresh after Spain beat France ↗ · Oddschecker — aggregated outright title odds ↗ · Silver Bulletin — Nate Silver's World Cup model ↗
Didn't the models just blow it?
France was everyone's favorite and it's out. A probability isn't a promise: Opta's last pre-semifinal run gave France 34% for the title and 57.7% to reach the final, so 66% said no title, 42.3% out here. The unlikely branch happened.
'Everyone predicted France' overcompresses: 34% was just the top number in a field where nobody cleared 50%.
Calibration catches bad forecasters: when they say 60%, does it happen 60% of the time? Each probability here carries a snapshot date and a post-final Brier score.
A day later, Opta leaned England (52.3%) in semifinal 2, and Argentina won. Two slim favorites down in two days isn't the models failing twice: near-coin-flips fall the other way constantly. I grade in bulk, not branch by branch.
Try it: the Monte Carlo bracket
Run 10,000 tournaments from the current bracket. Adjust each team's strength, or push the labelled bias nudge to see what interference looks like.
Speculative
Strength
Title odds
The math — Monte Carlo
Each match uses the logistic Elo formula P(A beats B) = 1 / (1 + 10^(−(Rᴀ−Rʙ)/400)), R being strength. A seeded generator draws each result; champions are tallied over 10,000 tournaments. At nudge = 0 the title odds sit near the clean model; the 'bias nudge' adds a capped few points to Argentina per match.
Soft, structural bias alone (H3) might add 2–5 points to Argentina, nudging title odds into the low-to-mid 20s. A strong H1 (unproven, unlikely on base rates) would show in tight knockout margins and could reach roughly 30–45%, a high end I doubt. A what-if, not a prediction.
Curious how this actually works — the formulas, the priors, the assumptions?
Not our year. And still, gracias
Argentina is out. It stings, but the joy of this run was real and no scoreboard takes it back. A fair exit is its own evidence: the tight margins that fed the whispers cut the other way. Gracias, muchachos.