Sports · Soccer · World Cup 2026
Rankings, by Round
I'm running my rating engine over the 2026 World Cup from scratch, one round at a time, as the results land. Right now it's mostly noise — every team level, the groups not yet connected — and it sharpens into signal as the bracket links the field. No seeds, no priors; only what's happened on the pitch so far moves a team.
The whole arc
Every team's results rating across the rounds played so far. Early on the lines sit in a tight, near-meaningless huddle; they fan out as more rounds land — that fanning is the rating turning noise into signal. A team's line stops where its tournament ends. Highlight any side:
Rating is openskill's conservative estimate (μ−3σ): teams start equal and only in-tournament play moves them, so early rounds sit low and undifferentiated by design.
Round by round
The standings as they stood after a given round — results rating (what the scoreboard says) alongside the performance rating (what the underlying play says), with movement against the round before. Only the rounds played so far are selectable; it opens on the round in play.
June 2026
July 2026
Highlighted days show the selected round. Bold days have matches; the dot marks today. 🏆 marks the final. Future rounds fill in as they're played.
Group Stage · Matchday 1
standings after Jun 14, 2026Barely any signal yet: with most sides one game in, the graph is a scatter of disconnected pairs and the order is close to a coin toss. That's the honest early state, not a bug — it sharpens as the bracket connects the groups. It's still settling as today's matches land.
Within-cluster only: every team has played about once, so the graph is eight disconnected pairs with no common opponents — there's almost nothing to compare yet.
| # | Team | Results | Performance | Move |
|---|---|---|---|---|
| 1 |  USA | 5.4 | 0.0 ·#4 | — |
| 2 |  Australia | 4.5 | 0.0 ·#5 | — |
| 3 |  Mexico | 4.5 | 0.8 ·#2 | — |
| 4 |  Korea Republic | 3.4 | 2.3 ·#1 | — |
| 5 |  Scotland | 3.4 | 0.0 ·#6 | — |
| 6 |  Bosnia and Herzegovina | 0.8 | 0.0 ·#7 | — |
| 7 |  Brazil | 0.8 | 0.0 ·#8 | — |
| 8 |  Canada | 0.8 | 0.0 ·#9 | — |
| 9 |  Morocco | 0.8 | 0.0 ·#10 | — |
| 10 |  Qatar | 0.8 | 0.0 ·#11 | — |
| 11 |  Switzerland | 0.8 | 0.0 ·#12 | — |
| 12 |  Czechia | -1.8 | -0.7 ·#16 | — |
| 13 |  Haiti | -1.8 | 0.0 ·#13 | — |
| 14 |  South Africa | -2.9 | 0.8 ·#3 | — |
| 15 |  Türkiye | -2.9 | 0.0 ·#14 | — |
| 16 |  Paraguay | -3.8 | 0.0 ·#15 | — |
How this is computed
Two ratings come off one openskill engine. The results rating is fed the actual result, the move scaled by a capped goal margin; the performance rating is fed an expected goal margin built from the box score (shots on target dominate; possession is ignored; saves are excluded as circular). Penalty-shootout knockouts count as the draws they were after 120 minutes. Teams are ranked by a conservative estimate (μ−3σ), which keeps barely-tested sides humble early.
Every team starts equal — no seeds, no priors — so the Matchday 1 order is close to meaningless, and that's the honest point, not a bug. During the group stage the played-match graph splits into separate clusters; cross-cluster ranks don't fully settle until the knockouts connect the field. The stat weights were learned across the 2018 and 2022 World Cups (128 matches). Every number here is derived from the match records at build time, never hand-typed.