← Visualizations Lluis Gasso

World Cup 2026 Final — Argentina 🇦🇷 vs Spain 🇪🇸

Monte Carlo simulation · Dixon-Coles bivariate Poisson model · 500,000 simulated finals · MetLife Stadium (neutral) · 19 July 2026

🇪🇸
Spain
55.7%
to lift the trophy · fair odds ≈ 1.80
🇦🇷
Argentina
44.3%
to lift the trophy · fair odds ≈ 2.25
Spain 55.7%Argentina 44.3%

The headline: this is close to a coin flip with Spain a slight favourite. Both are the two highest-rated teams on the planet, separated by only 32 Elo points. That thin margin — not the goals, not the flags — is what tilts the trophy toward Spain in the long run of simulations.

How the tournament resolves

72.8%
decided inside 90 minutes
27.2%
level after 90' → extra time
13.6%
go all the way to penalties
1.40 – 1.20
expected goals (ESP – ARG)

Regulation-only outcome: Spain win 39.8% · draw 29.5% · Argentina win 30.7%. Extra time and shootouts then split the draws, most of them near-evenly, which is why the final margin stays tight.

Most likely scorelines (after 90 minutes)

A 1-1 draw is the single most likely 90-minute result — fitting for two elite, well-matched sides — which is exactly why extra time and penalties carry real weight here.

Why this method

The best-validated approach for predicting football scores and results is a Poisson goal model, and the field's benchmark refinement is the Dixon-Coles (1997) model. It treats each team's goals as a Poisson process driven by attack/defence strength, then adds a correction for the well-documented fact that basic Poisson under-predicts tight low scores (0-0, 1-0, 1-1). This is the same family of models used by professional rating systems and betting markets.

Team strength here comes from World Football Elo ratings, the rating system that best separates international sides:

TeamElo (Jul 2026)World rankModel xG
🇪🇸 Spain2232#11.40
🇦🇷 Argentina2200#21.20

The 32-point gap gives Spain an Elo expected result of 0.546 at a neutral venue. Each team's expected goals (λ) were calibrated so the model's regulation win/draw/loss probabilities reproduce that Elo expectation — then the full knockout format (90' → 30' extra time → penalties) was simulated 500,000 times.

How robust is the edge?

Very. Re-running the model across a wide range of assumptions about how many goals the final produces barely moves the needle, because the win probability is anchored to the Elo gap rather than to the scoring rate:

Assumed total goalsSpain winArgentina win
2.2 (very tight)55.8%44.2%
2.455.7%44.3%
2.6 (base case)55.7%44.3%
2.855.7%44.3%
3.0 (open game)55.5%44.5%
Reading this honestly: a 55.7% / 44.3% split is a genuine toss-up with a slight lean, not a confident call. The entire edge rests on Spain being 32 Elo points above Argentina; if you think that gap is noise, the true number is 50/50. The model also can't see line-ups, injuries, red cards, or a Messi moment — it prices the average final, and any single match lives in the variance. Penalties, reached in ~14% of sims, are modelled as near-coin-flips by design.

Method: Dixon-Coles bivariate Poisson (ρ = −0.13), Elo-calibrated expected goals, 500,000 Monte Carlo simulations of regulation + extra time + shootout. Elo ratings via World Football Elo Ratings (eloratings.net / international-football.net), July 2026.

← Back to all visualizations