The Manchester Derby: The 60th-Minute Goal and the Two-Layer Offside Check Technology Has Not Reached

In the 60th minute of the Manchester derby, the ball settled in the net after a close-range finish. Referee Michael Oliver blew his whistle, raised his arm, and pointed toward an offside position. The stands split into two opposing emotional directions. Then the big screen showed the familiar three-dimensional frame: digitised skeletons, a blue horizontal line, the point of contact circled like a full stop. The semi-automated offside system completed its run within seconds. Erling Haaland was in a legal position. The goal was awarded, and in that moment almost nobody inside the stadium thought there was a second layer of checking being left empty.

Days later, news arrived from the refereeing bodies: two members of the match's VAR team — Matt Donohue and Blake Antrobus — were stood down. The professional refereeing oversight body, Pro Ref, concluded that a second player had been in an offside position and had influenced the play, and that the incident should have led to an on-field review rather than being closed out inside the VAR room. The goal itself stands in the record. But the story behind it has just opened a crack that I believe will be discussed for a long time.

I sat with that footage far longer than a news report requires. Because what went wrong here was not a line that was a few millimetres off. What went wrong was an implicit assumption: that once technology has answered the hardest geometrical question, the entire legal question has also been answered. Semi-automated offside technology determines position; it does not determine behaviour. Yet in operational practice, the two questions are routinely collapsed into one. That is the point I want to dissect in this piece, and it is also why I am giving it more time than the result of the match itself.

Before going deeper, a data note I am obliged to state. The initial material I have names the player in the offside position as Enzo Fernández. That name conflicts with the Manchester derby squad list, and Enzo Fernández is registered at a different club. I am keeping the structural analysis of the incident intact, because it does not depend on a specific identity, but I am separating out the identity question and flagging it as a detail requiring verification. In my trade, a wrong name can ruin a correct analysis, and I will not trade credibility for a few lines of apparent completeness. I refer to him as "the second player" throughout the remainder of this piece, until a source confirms otherwise.

The necessary context here consists of two pieces. The first is how semi-automated offside technology operates in English football. The system uses a dedicated network of cameras mounted under the stadium roof, capturing dozens of frames per second, building three-dimensional models of each player and identifying the precise moment the ball leaves the passer's foot. The result is an offside position reconstructed through geometry, no longer dependent on the viewing angle of whoever draws the line. In the majority of situations, this system delivers a faster and far more stable answer than the manual method. That is a genuine advance, and I have no intention of denying it.

The second piece is Law 11 — the offside law. There is a distinction here that most spectators do not need to know, but professionals are obliged to. The law distinguishes clearly between "being in an offside position" and "committing an offside offence". A player standing in an offside position has not necessarily committed an offence. He only commits one when he becomes involved in active play in one of the ways the law lists: obstructing the goalkeeper's line of sight, challenging an opponent for the ball, making an obvious action that clearly impacts an opponent's ability to play the ball, or gaining an advantage from that position. Interpreting these criteria belongs to human beings, not to geometry.

Put the two pieces together and you have a process that must necessarily pass through two layers. Layer one is objective, handled by machines. Layer two is subjective, and must be adjudicated by the on-field referee, and therefore must be returned to the pitch in the form of an on-field review. When layer one returns a "legal" verdict for the goalscorer, the VAR team's task is to continue into layer two with the remaining players in that passage of play. It is precisely this transition that was skipped in the derby.

I want to be clear about what I believe is the core issue here, because it is easily misread as a personal accusation. The operational failure in this incident is systemic rather than individual: the system has been designed to optimise the objective layer to such a degree that the subjective layer becomes ancillary work, handled through memory and experience rather than through a mandatory checklist. When a VAR team has just confirmed a complex offside line, there is a very real psychological pressure to believe the hardest part of the job is done. What remains looks like a minor detail. But in the laws of football, that minor detail is precisely where a goal is awarded or disallowed.

There is one technical aspect I consider more notable than all the rest: the semi-automated offside system has no function that alerts operators to the existence of a second offside player capable of influencing play. It draws the positional picture of every player, but it does not classify which of them is "involved in active play". That picture is accurate as data, and silent as law. For a VAR team under time pressure, an accurate but silent picture is easily read as a complete picture. Data does not announce what it is missing; the person reading the data has to announce it.

I have seen this mechanism repeat itself, at a much smaller scale, many times in my own analytical work. When you finish building a heat map of a team's activity zones, the satisfaction of seeing the image appear makes it easy to skip the question of what the model does not measure — the manager's intent, the players' fatigue in the 75th minute, the movement of the bench. Numbers have no gender, only correctly placed pressure. And the correctly placed pressure inside a VAR room is usually the pressure to finish a review that has already run too long.

The public part of the story is clear and easy to follow. Pro Ref concluded that the influencing offside should have led to an on-field review. Matt Donohue, who directly checked Michael Oliver's decision, and Blake Antrobus, the assistant VAR, were stood down. Their appointments will be reconsidered after the international break. Howard Webb, who heads the professional refereeing body, sits in a medium-to-high accountability pressure zone, because every error at this level is traced back to the operating standards he owns.

I do not regard standing officials down as a solution. It is a ritual of accountability, necessary but insufficient. Because it answers the question "who" and leaves the question "what" blank. If the process does not change, then another VAR team, in another matchweek, facing another double offside, will confront exactly the same gap. Standing down two people does not fill that gap.

What I want to place alongside that is an observation about the asymmetry of the process. In a goal situation, the VAR team is required to check every possibility that could lead to the goal being disallowed. That is a long checklist: ball out of play, a prior foul, offside, handball. In principle, the VAR team must complete the entire list. In practice, the order of execution tends to follow the ease of the evidence: what the machine can answer gets done first, what requires reasoning gets done later, and what appears already answered gets skipped. Double offside sits exactly at the intersection of those three characteristics, which makes it the single most missable category of incident in the whole system.

I have spent years watching matches in Spain and writing about them for readers here. During that time I learned one thing about how systems operate under pressure: they always optimise for what is easy to measure, and leave what is hard to measure to human beings with the least support. European football has invested heavily in sensors, cameras and models, but the bulk of that investment flows toward the objective layer, where the benefit can be proven in numbers. The subjective layer still relies on one referee standing in front of a screen, with a few seconds to judge intent and impact. Tactics are not magic; they are mathematics wearing a mask. But the offside law contains a part that cannot be reduced to mathematics, and that part is being left unattended.

There is one post-hoc question I think the league should answer publicly, because it determines whether this is an accident or a pattern. In how many goal situations this season has a VAR team confirmed offside through semi-automated technology but failed to continue reviewing other offside players in the same passage of play? If that number is small, this is an isolated error. If it is not small, we are talking about a structural gap, and standing down two individuals becomes a symbolic treatment of the problem.

Turning to the counter-intuitive angle, I think our biggest mistake in discussing this incident is placing technology in the dock. Technology did its job correctly. The semi-automated offside system determined Haaland was in a legal position, and that conclusion was right. The problem lies in the fact that we have implicitly granted technology an authority it was never designed to have: the authority to declare an incident fully reviewed. That is a confusion between accuracy and completeness.

These two concepts differ in nature. An offside line can be accurate to the millimetre and still entirely incomplete, because it only describes one person's position at one instant, while the law demands a judgement about that person's impact on everyone else. When we raise the accuracy of layer one very high, we inadvertently create the illusion that layer two has been handled along with it. This is the class of error I call an execution blind spot: the process is not wrong on paper, but the way human beings operate it under pressure is wrong. And execution blind spots are the hardest to fix, because they appear in no guidance document.

My second counter-intuitive angle is harder to hear. Standing down VAR officials after every major error may be producing a perverse effect: it encourages excessive caution at the objective layer and haste at the subjective layer. A referee who knows that a wrong line will bring public professional consequences will pour all his concentration into that line. Once the line is correct, the psychological burden lifts, and the remaining work is handled in a state of relief. The incentive mechanism is inadvertently allocating attention to the wrong place.

I am aware I am talking about things very far from a results report. But if you read this incident as a results story, you get one line: the goal was awarded, two officials were stood down. Read it as a process problem, and you get a question applicable to hundreds of other incidents across the season. Behind every data table are human beings sweating, and behind every three-dimensional frame too: a referee standing at a screen, with the clock running, trying to judge the intent of a player he can only see through twelve cameras.

There is a market connection I want to raise, though it is not part of the original story. At this point in the cycle, with clubs balancing budgets for the coming season, each position in the Premier League table carries a concrete cash value, typically a few million pounds of difference between tiers. A goal wrongly awarded in a derby is not only an emotional loss. It is a difference in prize-money flows, in a club's commercial value, and in the negotiating position of its players. The integrity of officiating, in the world's highest-revenue league, is a valuable asset. When that asset erodes, the cost does not sit in the VAR room.

This leads me to a remark about how we discuss refereeing errors in general. We tend to treat them as emotional events: anger, then forgetting, until next time. Another way to treat them is as operational data. Every error is a data point about where the process is thin. If the league published error categories periodically, rather than only individual controversial incidents, we would have a picture of patterns instead of a string of anecdotes. For a competition that invests in data to the point of tracking every touch of the ball, failing to publish data on its own errors is an inconsistency.

I have followed matches in La Liga and in England long enough to notice something about spectator reaction. Fans do not react with anger to technology. They react with anger to the feeling that there is a decision layer they are not allowed to see. When a three-dimensional frame appears on the big screen, spectators are shown evidence, and they believe. When a decision is reached with no evidence shown, they lose faith. In this derby incident, the strongest evidence was shown, and it was correct. But it was correct for half the incident, and the other half had no image at all. The visual silence at layer two was read as the absence of a problem. An empty stadium does not remove the noise; it only filters out what matters.

So what to do with that gap? I have three technical proposals, not criminal ones. First, add a mandatory step to the process: once semi-automated technology confirms the goalscorer is in a legal position, the VAR team must conduct a systematic sweep of all other attacking players inside the offside zone at the moment of the touch. That is an auditable step, and therefore a measurable one. Second, publish the VAR audio for double-offside incidents, similar to what has been done for other contentious situations. Third, when an incident contains a subjective element, the default must be an on-field review, not a VAR-room decision. Returning adjudication authority to the on-field referee does not slow the process down; it makes the process transparent.

I know these proposals sound dry. But they are verifiable, and that is the standard I set for everything I write. Over fourteen years covering this industry, I have learned that solutions which sound impressive are usually un-implementable, while solutions which are implementable usually sound very boring. Checklists are boring. Audits are boring. Publishing audio is boring. But those are the things that actually change human behaviour under pressure.

As for the clubs, I do not think there is much to say at this point. Manchester City benefited from the decision. Manchester United were disadvantaged. Neither club has the power to change the result. A formal complaint has value only as a historical footnote. What interests me more is the pressure this incident places on the next derbies, because matches with the highest scrutiny are always where processes are tested hardest. When a VAR team knows the whole country is watching, they will be more cautious at the objective layer and may still be hasty at the subjective layer, because the subjective layer is invisible.

And here is where I want to close with a forward-looking thought. People say I do not belong here, but data does not lie. The data in this derby incident says Haaland was legal, and it is right. It also says another player was in an offside position, and it is right too. Both truths coexist, and football needs a process capable of handling both at once rather than handling the easy one first and hoping the hard one disappears. In the next derby, when the big screen again shows those digitised skeletons, what is worth watching is not the blue line, but the silence after the line disappears. That silence is where the match is actually decided.