MACHINE LEARNING PREDICTION · FINAL RECORD

🏆 FIFA World Cup 2026

Live version (matches already played are fixed) · Poisson (Dixon-Coles) + Elo · 1,000,000 Monte Carlo simulations · the complete record of the 2026 World Cup

Validated on ~4500 unseen matches: 60% accuracy · RPS 0.172 (naive 0.228) — bookmaker level.

🏠

🏆

World Champion: Spain

Runner-up: Argentina

The final

10 ⚽

Golden Boot: Kylian Mbappé

🏆 The World Cup is done — open the full Tournament Review: how it played out & how the model did →

📅 Today's matches kickoffs in WEST (UTC+1)

📊 Odds over time since kickoff

Final bracket how the knockouts played out

The 12 groups final standings · who advanced

How the group stage finished. The figure on the right was the model's chance each team would reach the knockouts (it already accounted for the 8-best-third-placed-teams rule) — now settled to who actually went through: green = advanced, red = out.

🆚 Match Lab head-to-head

Pick any two of the 48 teams and see what the model makes of the match-up — win / draw / win odds, the likeliest scorelines, goal markets and more.

✅ Results so far predicted vs actual

😱 Biggest surprises model's worst calls

🏅 Model vs FIFA ranking does the model agree?

🎯 Does it actually work? track record

🏆 The 2026 World Cup: predicting live, the model called 71/104 match results right (home / draw / away) — and, blind, its top two pre-tournament picks were the two finalists and its top four the four semi-finalists (the champion, Spain, was its narrow #2). The historical track record below is how it does on tournaments it had never seen when it was trained.

The honest test: for every major tournament since 2002, the model is retrained on only what was known before that tournament kicked off and then predicts it blind — exactly how you'd have used it in real time. That's 2,757 matches across 67 editions (World Cups, Euros, Copa América, Nations League, Asian & African Cups, Gold Cups). Verdict: 56% of results called correctly (RPS 0.189) vs the naive base rate's 44% (0.231).

Competition	Matches	Model	Naive	RPS
UEFA Nations League	658	54%	43%	0.195
Africa Cup of Nations	493	51%	43%	0.189
FIFA World Cup	488	60%	44%	0.192
Gold Cup	340	64%	53%	0.175
UEFA Euro	246	50%	38%	0.207
AFC Asian Cup	230	61%	44%	0.180
Copa América	222	54%	45%	0.186
Confederations Cup	80	68%	44%	0.164

And zooming in on two tournaments the model had never seen when it was trained:

👟 Golden Boot top scorer

⚽ Anatomy of a goal every international goal

🟨 Discipline cards so far

📉 Elo through history drag the year

Year2026

🕰️ Historical facts 1872–2026

Biggest win ever: Australia 31-0 American Samoa (2001-04-11) · top international scorer: Cristiano Ronaldo (124 goals) · 49,520 matches, 145,570 goals.

History card — Spain (the current favourite)

468W-183D-140L (59.2%) · biggest win 13-0 vs Bulgaria (1933) · worst loss 1-7 vs Italy (1928) · longest unbeaten run 38 · rival Portugal (17W-18D-8L in 43) · peak Elo 2253 (2026)

Biggest upsets ever (underdog won, by Elo):

Year	Match	Tournament	Δelo
1980	Luxembourg 3-2 South Korea	Friendly	+770
2009	Bolivia 2-1 Brazil	FIFA World Cup qualifica	+536
2016	Georgia 1-0 Spain	Friendly	+526
2023	Kazakhstan 3-2 Denmark	UEFA Euro qualification	+522
2007	Equatorial Guinea 1-0 Cameroon	African Cup of Nations q	+520
2010	Niger 1-0 Egypt	African Cup of Nations q	+515

👟 Shooting first wins 52.3% of shootouts · penalty kings: Padania 100% · Indonesia 91% · Ethiopia 88% · Guinea 80%. ⚽ Goals/match in the 2020s: 2.72 · home wins in the 2020s: 51.1%.

⚔️ Beat the Machine you vs the model

The tournament's done — here's the final scoreboard. Each result you called scored points (✅ right result +10 · ⚽ each team's exact goals +5 · 📏 goal difference +5 · 🎯 exact score +5, perfect = 30), and the model played every match too. Sign in with Google to see your final scorecard, your match-by-match picks and where you finished on the leaderboard.

Data-driven model, just for fun. ⚽ Data: International football results 1872–2026.