Michael Carrick has had a rotten week. Manchester United lost 4-1 at Como on Wednesday, three days after a 1-1 draw at Leeds and eight days after a 2-0 defeat at Atlético Madrid, and they're 13th in the Premier League with 7 points from 7 games.
The awkward part is that none of it came out of nowhere. Since his first match in charge on 17 January, United have taken 46 points from 24 league games, third behind Manchester City and Arsenal over that stretch. On expected points, worked out from the chances each side created, they've earned about 36, which is seventh. That's roughly ten points more than the football deserved, and the gap was there from his first game.
Premier League, per game, 17 January to 18 October, ranked among 23 teams
| Per game | 24 games | Rank |
|---|---|---|
| Points | 1.92 | 3rd |
| Expected points | 1.51 | 7th |
| Goals for | 1.79 | 3rd |
| xG for | 1.69 | 6th |
| Goals against | 1.17 | 8th best |
| xG against | 1.39 | 7th best |
| Shots on target for | 5.5 | 1st |
The attack is genuine. United lead the league over the period on shots on target, 5.5 a game, and they're sixth for expected goals. The surplus is mostly at the other end. They've conceded 28 from 33.4 expected and scored 43 from 40.6, so of the 7.8 goals by which results beat the chances, 5.4 came at the back, in front of Senne Lammens.
They dominated the first game, a 2-0 home win over City, and were half a point up on expected points. By the third game they were nearly five up. After that the gap never dipped below zero, not once in 24 games.
Positive from the first game, and never below zero
| After | Points | xP | Gap |
|---|---|---|---|
| 1 game | 3 | 2.5 | +0.5 |
| 3 games | 9 | 4.1 | +4.9 |
| 6 games | 16 | 8.8 | +7.2 |
| 10 games | 23 | 15.1 | +7.9 |
| 17 games | 39 | 25.1 | +13.9 |
| 24 games | 46 | 36.1 | +9.9 |
Four of the 13 wins came in matches where the opponent created more: 3-2 at Arsenal (xG 0.7 to 1.9), 3-2 over Fulham (1.6 to 2.5), 1-0 at Chelsea (0.3 to 1.7) and 2-1 over Brentford (1.2 to 1.4). That's 12 points from four games they arguably shouldn't have won.
Look at who has been beating the chances. Matheus Cunha has seven goals from 2.9 expected, Benjamin Šeško eight from 4.7 in 668 minutes, and Casemiro five from 2.4, which is 20 goals from 9.9 expected between the three. Bruno Fernandes has 17 assists from 8.7 expected. The player with the biggest chances, Bryan Mbeumo (10.6 xG), has scored nine.
Premier League since 17 January: goals and assists against expected
| Player | G | xG | A | xA |
|---|---|---|---|---|
| Matheus Cunha | 7 | 2.9 | 6 | 2.3 |
| Benjamin Šeško | 8 | 4.7 | 0 | 0.1 |
| Casemiro | 5 | 2.4 | 1 | 1.6 |
| Bruno Fernandes | 7 | 6.4 | 17 | 8.7 |
| Bryan Mbeumo | 9 | 10.6 | 3 | 2.5 |
That's four players running well above the chances at the same time, and the one with the best chances is the one running under. The full list is on the Premier League attacking stats page and the United squad page.
In the 21 league games before his first match, United took 32 points, 1.52 a game, from 35.9 expected, with an expected goal difference of +0.65 a game. Since then it's 1.92 a game from 1.51 expected, with an expected goal difference of +0.30. Points per game went up by 0.4 while the underlying numbers more than halved.
It's not a perfect comparison, since the samples are small and the opponents differ, but every other number in the tenure points the same way.
A gap of 9.9 points is about 1.7 standard deviations from what the chances predicted (z=1.73, p=0.08). That's the kind of gap pure luck produces about one season in twelve, so it's suggestive without being proof.
Some of it could be real quality. Manchester City are +12.8 over the same stretch and Arsenal +7.5, because the best sides tend to beat expected points a little through finishing and goalkeeping that xG doesn't credit. United's chances rank seventh, though, a long way off those two.
Split the tenure and it's obvious. From January to May, 17 games, United scored 1.94 a game from 1.68 xG and conceded 1.06 from 1.48: 39 points from 25.1 expected, a surplus of 13.9. This season they've scored 1.43 from 1.71 and conceded 1.43 from 1.17, and have 7 points from 11.0, a deficit of 4.0.
The chances barely moved. Expected goals for were 1.68 a game and are 1.71; expected goals against were 1.48 and are 1.17. On expected goal difference, +0.54 a game, this is their best stretch under Carrick, and it's also the worst for results.
Premier League 2026/27, United first. The last column is points won minus points the chances earned
| Opponent | Score | xG | Pts v xP |
|---|---|---|---|
| Hull City (a) | 0-2 | 2.2-1.0 | -2.1 |
| Ipswich Town (h) | 5-2 | 4.4-0.9 | +0.2 |
| Everton (a) | 2-2 | 0.8-1.1 | -0.2 |
| Manchester City (h) | 0-1 | 0.8-0.8 | -1.4 |
| Fulham (a) | 1-1 | 2.1-1.5 | -0.8 |
| Tottenham (h) | 1-1 | 0.7-1.7 | +0.3 |
| Leeds United (a) | 1-1 | 0.9-1.3 | 0.0 |
Hull is the one that hurts: 2.2 xG to 1.0 and a 2-0 defeat. Manchester City at Old Trafford cost 1.4 points on its own.
That gives three versions of the same manager.
Points per game, and the pace that implies over 38 games
| Rate | Per game | 38-game pace |
|---|---|---|
| Tenure points | 1.92 | 73 |
| Tenure expected points | 1.51 | 57 |
| This season so far | 1.00 | 38 |
Our test of 40 Premier League team-seasons across 2024/25 and 2025/26 found that early gaps between points and expected points neither bounce back nor hold reliably (r=0.20, p of about 0.2), so the 73 and the 38 are both poor forecasts. The 57 is what the chances back.
The spring was built on results the football never matched, and this season is the bill. The expected goals for the Como and Atlético defeats aren't in yet, and seven more league games will double this season's sample. Both will say more than the scorelines did. The live numbers are on the Premier League xG table, the power rankings and the United team page, and our Como preview has the team news from before the game.
Data from our own match database, correct to 22 October 2026.