Spain 1–0 Argentina, aet. The model went into opening day with Spain as its top pick at 16% — above the market's 14.9% — and had Argentina, the eventual runner-up, at 11.9% against a market price of just 8.2% — its top two picks of the whole field were exactly the two teams that reached the final.
xWDL says the pick aged even better than the scoreline: Spain finished with the field's best expected-goal difference, was the xWDL favourite in all eight of its games, and conceded once all month — while actually under-finishing its own chances. The least lucky champion on the board.
The market won the tournament overall — but the model was dead even in the 32 knockout games, its whole deficit came from the opening third (thin priors on debutant squads), and the gap halves when both are scored on xWDL instead of the scoreboard.
Both forecasters beat the naive baselines by a distance (coin flip 1.099, season base rates 1.069). Match by match the market took 64 of the 100 duels to the model's 36 — the profile of an underdog-backer: frequent small losses, occasional bigger wins. Scored on post-match xWDL instead of the scoreboard, the gap halves (0.938 vs 0.923). Structurally the two disagree in one repeatable way: the market prices favourites +0.028 win-probability points above what xWDL says they deserve, the model -0.021 below — the model is the buy-the-underdog side of the market.
The model gave the world-cup debutants a draw price half again the market's — and xWDL says even that was too low.
The model gave Paraguay a 31% chance of taking the tie vs the market's 18.7% — and Paraguay took it.
The model made Belgium a clear favorite while the market leaned USA; xWDL says Belgium should have been priced higher still.
The market was sold on France; the model called it even — and xWDL says England deserved the ten-goal classic.
Both underrated DR Congo — but the model much less so, and xWDL says DR Congo were the better side on the day.
The model's biggest genuine miss: it had this near even while the market made DR Congo a clear favorite — and xWDL says they should have been favored even more heavily.
The model actually leaned Panama; market and xWDL both had Ghana as the stronger side.
Same weak spot: the model priced the eventual comfortable favorite as a coin flip. Underrating certain favorites was its one repeatable failure mode.
The mirror image of its underdog edge: the same lever that priced Paraguay and Cape Verde correctly occasionally shaved too much off a genuine favourite.
Mexico's perfect group came on 6.24 expected points — the luckiest run in the field. Turkey and Uruguay, the two unluckiest teams, were both eliminated.
Only 4 matches all tournament saw a side with a >65% xWDL win expectancy fail to win — all group-stage finishing failures, with Ecuador's 0–0 against Curaçao on 3.05 xG the most extreme. The biggest knockout deserved-result reversal: Brazil out-created Norway to a 0.602 xWDL win probability in the Round of 16 — and lost 1–2.
Beat Brazil — the market's #3 title pick — in the Round of 16, the tournament's cleanest shock, then took England to extra time.
Knocked out Germany in the Round of 32 on penalties — a bottom-third squad ending a top-8 favorite's tournament in the first knockout round.
Put out Netherlands on penalties in the Round of 32, then fell only to France.
A shootout win over Colombia carried it to the quarterfinals before Argentina ended the run — in extra time, naturally.
The model's #37-rated squad drew Spain 0–0, reached the knockouts, and took eventual runner-up Argentina to extra time before losing 3–2.
A top-8 favorite out in the first knockout round, to a team priced at 0.45%.
Same round, same script: eliminated on penalties by 1.7%-shot Morocco.
The market's #3 pick — the model was notably cooler at 5.0% — lost to Norway, the biggest market-rank-to-finish gap in the field.
One of the model's top pre-tournament picks lost its group to Colombia, then met Spain a round too early.
The model's Group D favorite at 39.2% to win the group finished dead last while co-host USA won it — the market had this one right.
For all the bracket chaos, 10 of 12 groups were still won by the model's pre-tournament favourite — the upsets came almost entirely in the knockout rounds.
| Player | Team | Goals | Min | G−xG |
|---|---|---|---|---|
| Kylian Mbappé | France | 10 | 699 | +3.12 |
| Lionel Messi | Argentina | 8 | 740 | +2.69 |
| Jude Bellingham | England | 7 | 616 | +3.94 |
| Erling Haaland | Norway | 7 | 465 | +2.56 |
| Harry Kane | England | 6 | 654 | +2.46 |
| Ousmane Dembélé | France | 6 | 597 | +3.94 |
| Mikel Oyarzabal | Spain | 5 | 606 | +1.00 |
Messi at 38: 8 goals, a tournament-best 4.15 expected assists, and 740 minutes — Argentina's entire engine, and the only regular on the squad who clearly outplayed his pre-tournament rating.
Spain's best-performing outfielder relative to expectation.
Modestly rated, quietly dominant in the engine room.
Spain's top scorer — a 0.65-quality squad player by the model's pre-tournament book.
Meanwhile Spain carried the model's two highest-rated goalkeepers in the entire 1,246-player pool — David Raya and Joan Garcia — for exactly 0 minutes between them. Emiliano Martínez played all 810 possible minutes: Argentina's road ran through three extra-time matches to Spain's one.
Rated near the bottom of the entire player pool pre-tournament; finished as one of its highest-output defenders.
The single largest rating-vs-output gap at the tournament — the anchor of the back line that took Argentina to extra time in the Round of 32.
The model's #2-rated player pre-tournament and the tournament's single biggest finishing underperformer; the chances came, the finishing never did.
Below-median tournament output — yet a champion, and remarkable for carrying starter minutes through a title run as a teenager.
96th-percentile rating in, single-digit-adjacent output out.
Played nearly every England minute without ever imposing himself on the numbers.
Want the receipts? The full match-by-match scoreboard is on the performance page, and the method page explains how the model works.