Erling Haaland was onside. That is the strange place to start the story of a goal that should have been disallowed.
Manchester City's winner against Manchester United has already become the biggest Premier League VAR error of the young season. The assistant referee raised his flag. VAR overturned the decision. Haaland's goal stood and City won 1-0. Pro Ref later admitted that was wrong.
Now we know much more about how it happened. Howard Webb has released the officials' conversation from the review and described what happened inside the VAR room as "tunnel vision". The phrase is revealing because the technology itself did its job: semi-automated offside established that Haaland was three centimetres onside.
The mistake came next. VAR became so focused on proving that Haaland was legal that it failed to properly judge what Enzo Fernandez, who really was in an offside position, was doing directly in front of Manchester United's defenders and goalkeeper.
That makes this more interesting than another argument about whether an offside line was drawn correctly. It is a case study in how VAR can answer the question it is given perfectly and still produce the wrong decision.
What actually happened with Haaland's goal?
Haaland turned Josko Gvardiol's cross into the Manchester United net in the 60th minute at Old Trafford. The assistant referee flagged for offside, so the on-field decision was no goal.
VAR Matt Donohue began the review by checking Haaland's position. Semi-automated offside technology established that Patrick Dorgu had played the striker onside by three centimetres. So far, everything worked.
The VAR then needed to check the full attacking possession phase before overturning the decision. That included Enzo Fernandez.
Fernandez was in an offside position in the middle of the penalty area. He launched himself towards the cross and narrowly failed to make contact before the ball reached Haaland. The VAR checked whether Fernandez had touched the ball. He had not, and the officials moved on.
That was the mistake. As Webb subsequently explained, an attacker does not have to touch the ball to commit an offside offence.
Fernandez's attempt to play it affected the defenders around him. Lisandro Martinez's ability to challenge for the cross was affected and goalkeeper Senne Lammens also had to react to the possibility that Fernandez would make contact.
Webb's conclusion was unambiguous: the goal should have been disallowed.
The VAR audio shows where the process broke
The released conversation is more useful than simply knowing that an error occurred because it shows the direction in which the review was moving.
The officials knew the original flag had been raised against Haaland. They therefore began by resolving Haaland's position. Once the technology showed him onside, the review effectively acquired a new assumption: this was probably a good goal.
The VAR said he still needed to check the full attacking possession phase, and he did. But when Fernandez appeared, the discussion centred on one question: did he touch the ball?
The replay showed a clear gap. No touch. Goal.
What was missing was the next question: Did Fernandez interfere with an opponent while in an offside position?
Under Law 11 of the Laws of the Game, touching the ball is only one way to commit an offside offence. A player can also be penalised for interfering with an opponent, including preventing an opponent from playing or being able to play the ball, or making an obvious action that clearly affects an opponent's ability to play it.
That was the part of the law the review failed to properly evaluate.
Howard Webb called it 'tunnel vision'
Webb's explanation is unusually candid. He said the VAR became "super-focused" on Haaland. Once Haaland had been established as onside, the officials did not fully evaluate Fernandez's impact. Webb described that narrowing of attention as tunnel vision.
There was supposed to be another safeguard. Assistant VAR Blake Antrobus was part of the team precisely because one official should be able to identify something another has missed. That did not happen either.
Webb said it was the responsibility of the rest of the team to pull the VAR out of that tunnel. Nobody did.
Both officials will now remain off Premier League duty until after the international break. Pro Ref has also contacted Manchester United, acknowledged the error and shown the club the audio and footage.
That level of transparency is welcome, but the explanation raises a difficult question.
What exactly is semi-automated offside solving?
Semi-automated offside technology is designed to make one part of the decision faster and more accurate: establishing where the players were. In this case, it answered that question correctly. Haaland was onside by three centimetres. Fernandez was offside.
The system could establish player positions. What it could not determine was whether Fernandez's actions constituted interference with an opponent. That required human judgement.
This distinction matters because "offside" is often spoken about as though it were now an entirely factual decision. It is not. The positional part can be factual; the involvement part sometimes is not.
A player who receives and touches the ball from an offside position presents a relatively simple problem. A player who never touches it can present a much harder one. Did he challenge an opponent? Did he obstruct someone's line of vision? Did an obvious action affect an opponent's ability to play the ball?
Those are football judgements. No automated line can make them disappear.
VarVote's guide to VAR intervention rules looks at the same divide. Technology can settle many factual questions extremely well. VAR becomes harder when the Laws require interpretation.
Old Trafford provided an unusually clear example.
VAR corrected the assistant and made the decision worse
There is another reason this error matters. Without VAR, the goal would not have counted because the assistant referee raised the flag.
The original decision was wrong about Haaland, but the final outcome would accidentally have been correct because Fernandez had committed the offside offence. VAR then introduced more accurate information: it correctly established that Haaland was onside.
By using that accurate information incompletely, however, VAR changed a no-goal decision into a goal that Pro Ref now says should never have stood.
That is an uncomfortable result for a system designed to correct clear mistakes. It does not prove that VAR makes football less accurate overall, but it does show why measuring VAR purely by the accuracy of its technology misses the point.
A technically perfect offside line is useless when the review around it is incomplete.
This was not the same as a debatable penalty
Most VAR arguments are not like this one. A defender puts an arm across an attacker. One referee sees normal contact. Another sees holding. Supporters watch six angles and still disagree. That is why the "clear and obvious" threshold exists.
The Manchester derby error was different. Pro Ref is not defending the final decision as one reasonable interpretation among several. It has said the goal should have been disallowed.
That makes this a process failure rather than simply another subjective refereeing disagreement.
VarVote's Weekend Review already identified the unusual contradiction at the centre of the incident: VAR correctly established that Haaland himself was onside, yet failed to properly assess the player who actually committed the offence.
The released audio now shows how that contradiction happened. The officials did not lack the pictures or the technology. They narrowed the review too early.
Could a checklist prevent another mistake?
That may be the most useful question to come out of the incident. Aviation, medicine and other high-pressure environments use structured checks partly because experienced people can still become fixated on the most obvious problem.
VAR already has a process. The officials knew they had to review the attacking possession phase, and the audio explicitly confirms that. But a process only protects against human error when each part is actually completed.
An offside review involving several attackers could require a simple sequence:
- Which attackers were in an offside position?
- Did any of them play or touch the ball?
- Did any interfere with an opponent?
- Did any gain an advantage from a rebound, deflection or save?
- Only then: final decision.
That is not a proposed change to Law 11. It is simply a way of making sure every part of the existing law has been considered before a decision is communicated.
Webb has already said Pro Ref will review which elements of the process failed. The Manchester derby suggests the answer may not require better cameras. It may require a better defence against human fixation.
The assistant VAR matters too
The role of the assistant VAR deserves attention here. One argument for having multiple officials in the video room is redundancy: people miss things, and a second person can challenge an assumption, notice another player or force the primary VAR to revisit an angle.
Webb specifically referred to that safeguard when discussing the derby. The VAR became focused on Haaland. The rest of the team was supposed to pull him back towards the wider picture. That did not happen.
It is therefore too simple to describe this as one official making one bad decision. The review system failed collectively.
That is more concerning, but it also gives Pro Ref something concrete to improve.
Transparency is valuable when it exposes failure
There is a temptation for refereeing organisations to communicate only when an explanation supports the decision made on the pitch. This case demonstrates why the opposite is more useful.
The Premier League has published the VAR process. Webb has explained the mistake. Supporters can hear how the officials reached the wrong conclusion.
That cannot undo the effect the decision had on the match. It does something else: it allows the discussion to move beyond "VAR is terrible" and towards the specific reason this VAR review failed.
That is progress.
UEFA has also moved towards greater transparency this season through its technical explanations of selected Champions League VAR decisions. More of this would help.
Supporters do not need officials to pretend every decision is defensible. Sometimes the most credible explanation is simply that they got it wrong.
The bigger VAR problem is no longer the camera
Football has spent years improving VAR technology. Offside lines became more sophisticated. Semi-automated systems arrived. More camera angles became available. Reviews became faster.
Those improvements matter, but the Manchester derby exposed the limit of that approach. Technology established the difficult measurement correctly to within centimetres. The wrong decision still followed.
VAR remains a system in which technology supplies information to human officials who must apply the Laws of the Game. That final part cannot be automated away every time.
And perhaps that is the real lesson from Old Trafford. The question is no longer simply whether VAR can see enough. It can.
The question is whether the people looking at all that information can reliably recognise what matters.
For three minutes in the Manchester derby, they were looking directly at the answer. They just stopped asking the right question.
The wider argument remains open in VarVote's global vote on whether football should keep VAR.