Carrick at Old Trafford: Ten Decisions and a Data Gap

In the 65th minute at Old Trafford, after Phil Foden's 23rd-minute red card, the Stretford End began chanting Benjamin Šeško's name. The Slovenian striker had been warming up since the start of the second half. He was still on the bench. Nineteen minutes later, as Šeško stepped onto the touchline, the score was 0-1. Erling Haaland had just scored for Manchester City.

I re-watched that footage four times, looking for a timestamp rather than a player's mistake. The delay between the moment a problem appears on the pitch and the moment the coaching staff responds to it. At elite level, that delay is usually shorter than one half of football. At Old Trafford, it stretched for nearly twenty minutes, and the price was a goal.

The xG shock at Hang Day turned me from a spectator into a data reader. Since that day, whenever a match makes me angry, I do not write immediately. I open the spreadsheet first.

This season, Manchester United under Michael Carrick have dropped nearly as many points as they lost across the entire previous campaign, in a far shorter window. I cross-checked that claim against four benchmark matches. A 0-2 defeat at Hull at the MKM Stadium. A 1-1 draw with Everton, conceding a 95th-minute equaliser. A loss to Manchester City through Haaland's goal while holding a man advantage. And a defeat to Brighton after leading by two goals — the first time Old Trafford had seen such a turnaround since the Tottenham match of September 2026. Brighton have now won at Old Trafford in four consecutive seasons. United have two league titles in nine years, and last season exited both domestic cups in the first round.

Set beside that run of results is a media product: a list titled "ten decisions that could make Manchester United pay." I read it the way I read any report of unknown origin — separating event from interpretation, keeping what can be verified, flagging what cannot. One thing must be said before the detail: this document supplies no process metrics. No xG, no xGA, no PPDA. It is an indictment of decision quality, running purely on event data. That does not make it worthless. It merely sets the limits of every conclusion drawn from it.

I grouped the ten decisions into three clusters, moving from structure to in-game management.

The first cluster is selection logic. Against an opponent fielding Oli McBurnie, Carrick started Ayden Heaven — a 19-year-old left-footed centre-back — on the left side, compensating for the absence of Lisandro Martínez. Prioritising a left-footed build-up profile can be sound in principle, but it becomes an error when it comes with too thin a matchup filter. McBurnie is a forward who uses his frame, contests aerial balls and leans on young defenders. Placing a 19-year-old centre-back at exactly that hotspot is a decision verifiable through results, and the results recorded it.

The second cluster is attacking structure. Patrick Dorgu was stationed on the left wing. The knock-on effect: Matheus Cunha was pushed to centre-forward, Bryan Mbeumo returned to the right flank. Pre-season ran the opposite way — Mbeumo played as a striker throughout preparation, while Cunha did his best work starting from the left. When a side already holds three left-wing specialists, piling another into that corridor pushes its most effective attackers out of their proven zones. The attacking structure was scrambled before the ball was even kicked.

I have seen this in the V-League, and it is never pretty. A team stacks three players of the same specialism into one corridor, then wonders why the middle is empty. Football does not punish anyone; it silently records that you poured resources where they were not needed and left the places that were needed bare.

The third cluster is game management, and this is where I stop longest. Foden was sent off in the 23rd minute. United had 22 minutes of the first half with a man advantage to create separation. They created none. The opponent set up a deep three-centre-back block, and United lacked a focal point up front. Šeško — the striker profile built for exactly that problem — was only introduced after Haaland had scored.

Then came Everton. Dalot had been booked and was withdrawn. Leny Yoro came on, a centre-back playing narrow, exposing space for Jack Grealish. From that space the ball travelled to Dewsbury-Hall. Maitland-Niles' 95th-minute equaliser followed a hesitant challenge from Noussair Mazraoui, fearful of conceding a penalty. A chain of reactive substitutions degraded the defensive structure at precisely the moment that structure most needed to hold.

And Bruno Fernandes. He was pushed deeper to hand the No.10 role to Cunha, while United were trailing with more than twenty minutes left. The clearest late chance — the 95th minute, Mbeumo — arrived when Bruno had returned to the No.10 zone. Kazan does not take revenge; Kazan simply builds a table and waits for me to miscalculate. Here the table says one simple thing: pulling Bruno away from danger zones when a team needs a goal runs against the state of the match.

Then Joshua Zirkzee. The club were ready to sell him in the summer, Everton made contact on deadline day, and United kept him because the squad needed another body. Since scoring in Ruben Amorim's first Premier League win on 1 December 2026, Zirkzee has added only two more league goals. Keeping a striker you had already decided to sell purely to fill a quota reflects a recruitment cycle out of phase — a planning gap rather than a panic buy.

There is another layer the indictment only brushes past. Ayden Heaven is 19, Leny Yoro is still young, and a 15-year-old sat on the bench. Elsewhere, 15-year-old JJ Gabriel reportedly told the club he wanted to leave. Throwing academy defenders into elite fixtures while a young talent wants out is a signal about the credibility of the pathway from academy to first team. Marcus Rashford left the pitch after requesting an energy gel, and Dorgu replacing him was described as a negative signal to a senior attacker. Together these details sketch a dressing room losing faith in the model.

What gives the indictment weight is not any single decision but the repetition. All three clusters circle one idea: an effective operating template existed in the second half of last season, and this season the team is drifting from it. When a coach drifts from a formula that once worked, supporters forgive far more slowly than when he is building something new. I call it the regression gap — a team does not get worse for lacking ideas, but for abandoning the old idea before a replacement exists.

Within the wider league picture, Manchester City remain the benchmark at title level. Brighton have won at Old Trafford four seasons running, and Everton — with Grealish in the side — are closing the gap in the upper group. Clubs once cast as disruptors now occupy exactly the space United used to control. This is the point I always stress when writing about the small-town-beats-big-club story: the romantic narrative hides an operational gap. At Old Trafford, that gap is now reversed.

Operationally, this is the kind of decision I always flag in red. A club willing to sell a striker but unable to sign a replacement usually implies budget or wage constraints. I have no data to assert that at Old Trafford, so I leave it as a low-confidence hypothesis. What I can assert is the consequence: retaining a player you no longer want, then depending on him, weakens your negotiating position in the next window.

At this point I have to stop myself, because that is what I do whenever an indictment reads too smoothly.

Every argument above is an argument about decision quality, not about variance. Without xG, I cannot run the test that results will regress to the mean. Nor can I rule out the opposite — that some of those late concessions were noise, a single opponent shot finding the top corner. The source material gives me no tool to separate the two hypotheses, and I will not pretend otherwise.

One point matters more than the error term. The defeats described here — a 95th-minute equaliser, a collapse from two goals up — are psychologically corrosive. They feed a bottling narrative that outlasts a single result. When a team loses because the opponent is better, fans accept it. When a team loses because of its own changes, patience evaporates far faster. Belief is a noise variable; run the emotional regression before placing the bet. At Old Trafford right now, that variable is skewed hard toward impatience.

The second issue is the source. The analysis I cross-checked shows signs of unreliability: unknown origin, several self-contradicting timelines. I keep it because the specific decisions inside it can be verified from footage. I do not keep it as a verdict.

The media loop — frustrated fans, decision lists shared, chanting shaping selection — is a self-reinforcing cycle. It does not create results, but it shortens a coach's time. In my trade, that is the hardest risk to price, because it sits in no table of numbers.

So what can be replicated?

If the self-inflicted-wound hypothesis holds, it will fade quickly once the original variables are restored: Cunha back to the left, Mbeumo to the right or up front, Bruno pushed higher when a goal is needed, an experienced centre-back pairing shielding the young defenders. That is a testable experiment with a time frame — the next three to five fixtures.

I do not predict the future; I only read ahead at how the past keeps operating. Three years ago in the Bundesliga, when the stands were empty and the home-advantage coefficient in my model collapsed, I rewrote the entire system within 72 hours. The lesson was not how far home advantage fell. It was this: a broken model is the day the data monk must burn his book and start again from the root text.

The crowd leaves, the model breaks, and I learn to hear the breathing of an empty stand. At Old Trafford, the stands are not empty yet. But if this run of results extends a few more rounds, there will be a silence long enough for people to hear it.