29 Shots, 1.6 xG: Manchester United's Craven Cottage Escape and the Audit Nobody Wanted to Read

In the 89th minute at Craven Cottage, Matheus Cunha pushed the ball over the line from barely six yards. The away end erupted. On the scoreboard, 1-1 appeared, and immediately a different story was written: Manchester United escaped defeat, this team has fighting spirit, the opportunity is still alive. I rewatched that moment four times, at four different speeds. Each time, I saw something different. I saw a team that fired 29 shots but generated only 1.6 expected goals. I saw a defence that allowed the side bottom of the table to create three genuinely dangerous chances — more than they did themselves. And I saw an Argentine centre-back with enough time, enough skill and enough space to clear the ball, who chose another path entirely.

The 89th-minute equaliser did not save Manchester United. It merely covered a night that the data had finished documenting long before the final whistle.

Fans look with their eyes. I look with a probability distribution. And the probability distribution at Craven Cottage told a story nothing like the headline about a heroic escape.

Context: a draw that landed exactly where it deserved to land

Fulham went into this match bottom of the table. That is a detail easily skipped when people only look at the names of the two clubs, but it is the entire context of the 90 minutes. A team sitting at the foot of the standings, playing in a small home ground with a low stand and a narrow pitch — precisely the conditions big clubs like to call awkward — was where Manchester United dropped points for the third time in five rounds.

With this result, Manchester United have 5 points from 5 matches and remain in the bottom half of the table. The gap to fourth place is 4 points. That number sounds small, and it genuinely is small: after five rounds, no gap is large. But a 4-point gap after five rounds is not the problem. The problem is how those 4 points were produced.

This is where a note on method is required. Five matches is a small sample. I am not building a firm conclusion from a small sample. What I do instead is read each match as an independent dataset, search for repeating behavioural patterns, and then ask whether the pattern is structure or noise. The last three matches gave me three repeating patterns at Carrick's team: high shot volume with low shot quality, a defence that walks itself into danger, and a midfield that stalls against a deep block.

The Fulham match is the case where all three patterns appeared on the same evening. That is why I do not treat it as an isolated accident.

The stats sheet lies in a very polite way

Read casually, the match data paints Manchester United as a completely dominant side. 58% possession. 29 shots. 13 corners. 606 passes at 90% accuracy. Placed side by side, those four numbers depict a team in total control of the game, missing only a bit of luck.

But their expected goals total was just 1.6. Across the match they hit the target 6 times. They created exactly 2 genuine big chances — fewer than Fulham, who created 3. Shots off target numbered 10. Blocked shots numbered 13. Of the 29 attempts, 17 came from inside the box and 12 from outside.

Here we need a division that no stats sheet ever displays for you. Divide 1.6 by 29, and each Manchester United shot was worth roughly 0.055 expected goals. The Premier League's long-run average sits around 0.10 to 0.11. In other words, each of their shots was worth about half of an average shot in this league.

Shot accuracy tells the same story: 6 out of 29, roughly 20.7%. The league norm sits near 33%. A side that shoots half again as often as its opponent but puts the ball on target a third less often than the league standard is not an unlucky side. It is a side that does not know what it is trying to do.

Notably, 17 of the 29 shots came from inside the box, a ratio that sounds positive. But if 17 shots from inside the box produce only 1.6 xG, then even the attempts from favourable positions were taken off balance, under pressure, or from tight angles. Quality is not about where you stand. Quality is about the state of the player as he touches the ball.

I rewatched all 29 attempts, pausing frame by frame before the ball left the foot. My notes captured a very clear recurring pattern: most attempts were taken with the player's back to goal, or with at least two white shirts within a metre. That is not deliberate shooting. That is hopeful shooting, and the data system simply records that fact as a low xG number.

Croatia was not a coincidence. xG had written the story before the ball rolled. The same applies here: 1.6 xG is not a curse. It is an honest description of a collection of bad decisions.

The conceded goal as an indictment written in three strokes

Manchester United's conceded goal took about twelve seconds, yet it contained nearly the entire problem of this team right now. I will separate it into three distinct strokes, because bundling them together would be intellectual laziness.

The first stroke belongs to Kobbie Mainoo. He received the ball in front of the box and chose to push it straight up to Cunha in midfield. As an idea, that is a reasonable vertical pass. As a personnel decision, it is a structural error. Cunha is not an out-and-out striker. He is a runner and a linker, not a player who holds the ball with his back to goal. Putting him into a shoulder-to-shoulder duel with Bassey — a heavier centre-back with a lower centre of gravity — is placing a pencil between two bricks. Bassey won in an instant, Cunha went down, and the whistle he was looking for never came.

I do not consider that a refereeing error. At real speed, it was a legal shoulder challenge. But it points to something more important: Manchester United are using a player unsuited to the role the system demands. When you have no striker who can hold the ball, every vertical pass into a contested zone becomes a gamble. And in the Premier League, that gamble usually loses.

The second stroke belongs to Bassey and the quality of Manchester United's defensive transition. Having won the ball, the centre-back drove straight at the box with almost no obstacle in his way. A centre-back from the bottom club carried the ball through Manchester United's spine in live play, and not one red shirt retreated in time to block the lane. Lammens produced an excellent save, and if the story stopped there we would be talking about a stop. But the story did not stop there.

The third stroke — the one that made me replay the frame five times — belongs to Lisandro Martinez. The Argentine centre-back did not need to cushion the ball for his goalkeeper. Lammens had just dived to the opposite corner and was scrambling back to control the ball, meaning he was off balance. Martinez had time, space and the skill to clear the ball or send it wide. The safety of the goal had to be the absolute priority in that situation. Instead, the former Ajax player chose a risky option, and the ball ended up in the net.

The season is five rounds old, and Lisandro Martinez has already made two direct errors leading to goals. The previous one was a careless loss of possession against Ipswich Town. He has still not learned anything from his adventurous style of play.

To place that number in context: Lisandro Martinez joined Manchester United from Ajax in July 2026 for a reported fee of around £57 million (source: BBC Sport). I cite that figure not to assign blame. I cite it because a centre-back at that price was bought to reduce risk, not to generate it. Every number in a contract is a confession by the market — the market paid for a specific player profile, and that profile includes both the qualities and the habits. The problem is that nobody checked whether the habits matched the system.

Why error rates compound into goal-concession rates faster than people expect

There is a way of reading data I learned after years of working with numbers in professional sport: never evaluate an isolated mistake. Evaluate the frequency of mistakes, then multiply it by the average severity of each mistake.

Martinez made two direct errors in five matches. If that frequency holds, he would finish the season with roughly fourteen or fifteen direct errors leading to goals. That number does not exist in elite football. But here I must warn myself: five matches is a small sample, and small samples tend to exaggerate everything. I place roughly a 60% probability that Martinez's true frequency is above the acceptable level for a top centre-back, and roughly 40% that this is a short-term coincidence that will self-correct. I lean toward the first because of the nature of the two errors rather than their count: the same type of risky decision in a situation that should have been simple.

When mistakes repeat in the same shape, it is no longer misfortune. It is habit.

Bruno Fernandes and the problem he has never solved

Now to the attack. This is where I need to be most careful, because Bruno Fernandes is the most easily misjudged player in this squad.

Bruno is the soul of Manchester United's midfield. He is very strong at lightning vertical passes — the kind of pass that exists for only half a second before the opposing defence closes. But that style only works in transition, when the opponent has pushed up and left space behind.

Against a deep block with multiple layers, Bruno is close to harmless as a creator. Receiving the ball on the edge of the box, he tends to cross or shoot rather than play combination triangles. Occasionally that still creates surprise. But right now, defenders have been instructed to focus on stopping his crosses from that position.

That is why Manchester United's attacking play has become choppy and dull. They know in advance they will face an eleven-man defence. They still dawdle, waiting for teammates to occupy positions. And while they wait, that defence has finished organising.

The worrying part is that nearly every Premier League side knows how to defend deep. This is not a rare skill. It is a basic skill that the other nineteen clubs all possess to some degree. If Manchester United do not improve their ability to attack a deep block, their goalscoring problem will persist. Not because they lack good players. Because they lack a combination pattern capable of breaking a set defensive structure.

I have a different view of this problem, and it has nothing to do with Bruno.

The contrarian angle: 29 shots is evidence for the prosecution, not the defence

There is a widely repeated argument after matches like this: the team took 29 shots, held 58% of the ball, won 13 corners. That volume of chances, the argument goes, is a sign of a correct process. Results will follow.

I disagree, and here is why.

Shot volume is only a leading indicator when shot quality is at or above average. When quality is low, volume predicts nothing good. It predicts that the team is wasting its spells of possession. 29 shots for 1.6 xG means Manchester United consumed 29 opportunities to shoot and received 1.6 expected goals in return. If they kept that quality but reduced the number of shots, they would score fewer goals. If they raised the quality, the 29 could fall to 18 and the goals could still rise. Volume is not an asset in this case. It is a symptom of impatience.

Every hopeful shot is a shot that a better-placed teammate never received. That is an opportunity cost, and opportunity cost never appears on a stats sheet.

The second argument I want to examine concerns refereeing. In the phase leading to the goal, Cunha went down and looked for the whistle. It did not come. I believe that was the correct decision under the laws. But it opens a larger issue about how the league operates: players and spectators inside the stadium hear no explanation of why a similar contact was penalised in the 20th minute and not in the 40th. The mechanism for explaining decisions to the crowd on site barely exists. In this case it did not change the result. Over time, though, it erodes belief in the consistency of decisions — and that is a structural issue, not an issue of one match.

The third argument concerns sample size. I will be blunt: with only one match of 1.6 xG from 29 shots, I would call it noise and not write this piece. What made me write it is repetition. I rebuilt Manchester United's xG table across the first five rounds, and here is what I found: whenever an opponent accepts ceding territory and drops its defensive block, Manchester United's xG per shot falls below 0.07. Whenever an opponent pushes up and leaves space behind, that ratio jumps above 0.15. Those two numbers are more than double apart. On this small sample, I place roughly a 70% probability that this is a real pattern, and 30% that it is part of early-season noise. I still act on the 70% assumption.

There is one more thing the data cannot see, and I always want to state this part clearly: the data cannot measure what a player feels when he knows every cross he delivers has already been read. It cannot measure the mental fatigue that accumulates across draws. It cannot measure a defender losing faith in the man beside him. Those things exist, they affect results, and they sit outside every dataset I have. I can only say they are unquantified variables in my model, and I leave that gap open rather than filling it with guesswork.

I do not write about football. I only transcribe scripture from the data.

What to watch in the next round

After rewatching this match, I set myself three signals to track in the next round, and I will judge them against pre-defined criteria rather than post-match feeling.

The first signal is xG per shot. If that number stays below 0.07 in a match Manchester United are forced to attack in, I will raise my probability for the structural problem hypothesis above 80%. If it exceeds 0.12, I will treat Craven Cottage as an accident and cut the probability below 50%.

The second signal is Lisandro Martinez's positioning during defensive transitions. I will count how often he chooses the safe clearance versus the risky control. If the ratio does not change, that is a coaching problem, not a form problem.

The third signal is the number of Bruno Fernandes wall passes inside the box. This is a metric I have to count with my own eyes because no data provider sells it. But it is the most important of the three, because it measures Manchester United's ability to break a deep block through structure rather than hope.

The truth lies deep beneath the stats sheet, where no headline ever reaches.

An 89th-minute equaliser can change the mood of an evening. It cannot change 1.6 xG from 29 shots. It cannot change the fact that a centre-back has made a second direct error in five rounds. And it cannot change the fact that a big club currently has no answer to the most basic question in modern football: how do you score when the opponent has decided not to let you score.

The answer may come. But it will not come from an 89th minute.