Carlo Ancelotti has managed 799 competitive games in our database across his last nine jobs: Milan, Chelsea, Paris, two spells at Real Madrid, Bayern, Napoli, Everton and now Brazil. Set aside the Everton spell, the one job on that list outside the elite, and his teams finished a game with 33.6% of the ball or less 6 times in 720 club matches. The opponents who managed it: Bayern Munich, Barcelona, Manchester City and Liverpool. That is the entire list.
On Sunday night at MetLife, in front of 80,663 people, Norway joined it. And then they won the game. Saturday in Miami, against England, is where we find out what it meant.
The shortest list in his career
Brazil finished the round of 16 with 33.6% possession. That is their lowest share in all 141 senior internationals in our database, which reaches back to 2002. The previous floor was 36.7%, against Chile in 2015. Prime Spain never pushed Brazil this far down. Argentina never did. A Norway side that lost 4-1 to France three weeks ago and rotated ten starters to do it now owns a mark that belonged to nobody.
It is also new for this Brazil. Under Ancelotti they had played 10 competitive matches before Sunday and never dropped below 46.1% of the ball, averaging 56%. The Norway game sits 12.5 points under his Brazil floor. Whatever this was, it was an outlier even by the standards of a coach who treats possession as optional.
He rarely loses these games
Here is the complication, and the article is dishonest without it: the low-possession game is Ancelotti’s signature move, and he almost never loses it: one defeat in those six games. The most famous result of his career, the 4-0 at Bayern in the 2014 Champions League semifinal, came with 30.7% of the ball. Ceding possession and striking through the space it opens is a thing his teams do to opponents on purpose. So the fair question about Sunday is whether Norway took the ball or were handed it.
The chance count says the plan worked either way. Brazil out-shot Norway 14 to 9 and out-created them 2.16 expected goals to 1.03 on our shot model. That is the 2014 shape almost exactly. What our data cannot see is intent, whether Brazil sat back by design or because Norway’s midfield pressed them off the ball, and we won’t pretend otherwise. What it can see is the outcome: Brazil played an Ancelotti game, created chances worth double Norway’s, and went home because the finishing step failed. That step is the one part of the plan no coach has ever controlled.
The Haaland ledger
Norway’s tournament reads as a team finishing hot. It is more precise to say one man is. Norway have scored 12 goals from chances our model prices at 9.2 expected. Erling Haaland alone has 7 of them from 18 shots worth 4.0, none from the penalty spot. He is converting 39% of his shots where the locations say 22%. The rest of the squad, combined, sits at 4 player goals from 5.1 expected, slightly under water. Both goals against Brazil were his, a header worth 0.20 in the 79th minute and a strike from outside the box worth 0.05 in the 90th, both assisted by Andreas Schjelderup.
The ball-holding is also newer than it looks. Norway’s possession share in their first four games ran between 41.7% (against Senegal) and 61.3%; the 66.4% against Brazil was their own tournament high. So England prepare for a side whose control game has appeared once at full volume, on the biggest night available, and whose goals flow through one finisher beating his numbers. Both are waiting for England in Miami.
What the model says about Saturday
Before the tournament our simulation gave Norway a 16.0% chance of reaching this quarterfinal and a 1.4% title chance. Before Sunday’s kickoff it gave them roughly 25.4% to beat Brazil in ninety minutes. They are in the quarterfinal, so the model has been below the field on this team twice. Its current read: England 65.7% to reach the semifinal, Norway 34.3%, with Norway’s title chance up to 4.7%. The model never sees possession, and it never sees an individual finisher; it prices team-level chance creation, where England still project ahead. Whether that 34% is the third miss or the correction is precisely what Saturday answers.
The live test writes itself. If England dominate the ball in Miami and Norway sit where Brazil’s chances came from, the round of 16 was a coin landing heads for the second time. If Norway hold the ball against England the way they held it against Brazil, and Haaland keeps converting chances at double what the locations say they’re worth, the model is mispricing something structural, and next Wednesday’s semifinal preview will have to say so.
Ancelotti’s spells come from our coach-career table joined to our match database: competitive men’s games only, women’s sides that share a team name excluded, preseason friendlies excluded, and possession coverage starts in 2007, so his earliest jobs are underrepresented. The Everton spell stays in the chart, dimmed, for honesty: a mid-table side ceding the ball is a different phenomenon from a super-club doing it, and with it the list grows by 7 games. Expected goals are our location-based shot model; it values where a shot was taken from and does not see defenders or the goalkeeper. Matchup probabilities are a frozen 2026-07-09 snapshot of our Monte-Carlo tournament simulation; the live board carries the current numbers until kickoff.