The same 3-4-2-1, measured the same way at three clubs: his last full season at Sporting, fourteen months at Old Trafford and the first five Serie A games at Milan. What stayed identical, what bent to the players, and where the idea is already visible at Milanello.
The pitch on the right is not a diagram. It is the average position of every role, second by second, across 45 Premier League games at United, cycling through the six moments of his game. Pick a club or a moment to hold it.
Each cell is the team's value and its rank in that league and season (four sets of 18 or 20 teams, rank 1 always meaning "most" of what the row describes unless the note says otherwise). The same definitions run over every club, so the columns compare directly.
Two rows barely move across four seasons and three countries: the share of crosses delivered by a wing-back and the share of final-third receptions that land on one. First, second, third, first. Second, third, second, second. Whoever plays there, the attack is built around the wide man in a back five.
The build-up rows are where the idea bent. At Sporting the goal kick went short 70% of the time, top of the league, and the centre-backs moved the ball forward 25 times a match, also top. At United the short goal kick fell to 16th and 15th, and the centre-backs' forward actions to 4th and 5th. Milan are in between after five games.
The base is a back three with two holding midfielders in front, the keeper deep behind them, and the wing-backs starting almost on the touchline. In United's second season a clean line of three sat in front of the keeper in own-third possession more often than in any set we measured, and the keeper himself stood deeper: 7.3 m off his line on average, against 11.5 m in the months before Amorim arrived.
At Milan the back three is wider than it ever was at United (30.7 m between the outside centre-backs, against 25 to 29 m) and the wing-backs are pinned wider still: 70% of the time within 10 m of the touchline during the first phase, against 56% at United. The clip is the pattern in its cleanest form against Torino: a short goal kick, Gila wide right, into the pivot, out to the left half-space and a wing-back run in behind for the cut-back.
In the final third the shape splits cleanly into lines: the back three high at the halfway line and beyond, the two midfielders ten metres in front of them, and a front five made of the wing-backs, the two 10s and the 9. The wing-backs finish level with the 10s, which is why they deliver between a third and a half of the crosses and take a quarter of the final-third receptions at every club.
What changed between his two United seasons is the commitment behind the ball. With the ball in the final third, United kept 5.4 players more than 10 m behind it in 2024/25 and 4.8 in 2025/26 (5.9 before he arrived). Milan are already at 4.2. The Sunderland goal is the last 35 seconds of a one-minute move: build from the back three, the ball worked right, the wing-back high, the 10 arriving from the other side of the box.
The average spot of every open-play action per role, in games he started with a back three. The same skeleton at all three clubs; at Sporting everything sits about five metres higher, and the 9 lived eleven metres further up than United's.
There is no positional data for his Sporting side, so this is rebuilt from the events alone: the ball's path and the player on it. It is the same left-sided move as the average picture above: the outside centre-back carries into midfield, the wing-back takes it on the line, the 10 goes to the byline and cuts it back for Gyökeres.
When the opponent has the ball in their own third, the shape is a 3-4-3 more than a 5-2-3: the near-side wing-back steps out onto their full-back, at least 10 m ahead of the back three in half of those moments, while the far-side one stays closer to the line of three. The back three holds its line near the halfway line.
That took a season to install at United. The share of near-side jumps went from 44% to 51%, the back line during the opponent's build-up rose from 42.6 m to 48.8 m from goal, and the team shortened from 40 m to 34 m. In the second season United won the ball back in the final 40% more often than any team in the Premier League, and recovered it within five seconds of losing it in the opponent's half more often than anyone else too. Milan's wing-backs jump as often as United's did in year two (54%), with the line at the same 49 m.
Sporting were third in Portugal for shots within 15 seconds of an own-half regain. At United that fell to 20th in the first season, and 12th in the second, with the forwards chosen for the press more than for the run. Milan are 7th after five games, and the opener in Turin is the template: a regain 34 m from Milan's goal, three forward passes and a finish 13 seconds later.
At Sporting the same idea came from the back: Inácio wins it, Catamo takes it from the wing-back position and plays straight into Gyökeres, who sets Pote.
The price of a front five is paid at the moment of the loss. Two seconds after United lost the ball in 2025/26 they had 6.0 outfield players goal-side of it, the fewest of any set we measured (6.8 before Amorim, 6.8 at Milan now). When the wing-back is up and the outside centre-back is drawn to the ball, the channel between them is the space opponents aim for.
The Villa goal in December is the textbook case, and the dashed line in the clip measures it live: the distance between United's right centre-back and right wing-back as McGinn looks up, and where Rogers arrives.
It is worth being precise about the size of the problem, because the process numbers were good. United's second season was second in the league for expected goals created and fourth-best for expected goals against, while conceding the fourth-most goals and sitting sixth on points per game. The shape was producing the chances it is designed to produce; the conversion at both ends was not.
Five matches is a small sample, so read ranks as direction rather than verdict. Still, the shape is already there in the positional data and some numbers have moved a long way from last season's Milan.
Possession is up to 60% (3rd), from 52% under Allegri. The press is more intense (PPDA from 14.4, third most passive in Serie A last season, to 11.3). The wing-backs are the widest of his three teams in the first phase, and the rest defence is the most committed. What is not there yet is the payoff of the press: regains in the final 40% are 17th of 20, and winning it back within five seconds is 13th. At United, that part took a full season to arrive.
Every number on this page comes out of the same definitions, so it can run automatically after each Milan match and for every opponent: wing-back jumps, line height, rest defence at the moment of the loss, the channel distance, with the clips attached. Happy to walk through it, or to build the next one around the questions your staff are working on.