Minute 34, 30 September 2026, Tottenham Hotspur Stadium. Luis Díaz broke through, rounded goalkeeper Guglielmo Vicario and put the ball in the net. The assistant referee raised his flag. Referee Simon Hooper signalled: no goal. More than three hundred kilometres away, in the VAR room, Darren England and Dan Cook sat in front of their screens, reviewed the phase, and said two words into the intercom: "Check complete."

Four days later, on 3 October 2026, PGMOL released the audio of that exchange. Listeners heard a silence lasting several seconds after those two words. The silence did not belong to calm. It belonged to a process that had just collapsed without anyone in the room knowing.

In the same match, the VAR team did one very difficult thing correctly: they upgraded Curtis Jones's yellow card to red in the 26th minute, after reviewing his challenge on Yves Bissouma. One system. One match. Two opposite outcomes. Tottenham won 2-1 through Joel Matip's stoppage-time own goal; Liverpool finished with nine men after Diogo Jota collected two yellow cards.

People usually fold both incidents into one word: "VAR." And that folding is the root of almost every argument.

The legal foundation: a clause written by eight votes

VAR was born at IFAB — the International Football Association Board, founded in 1886 and the only body in the world with the power to amend the Laws of the Game.

IFAB's voting structure deserves close attention. The board has five members: the football associations of England, Scotland, Wales and Northern Ireland hold one vote each, and FIFA holds four. Eight votes in total. To change a law, a proposal needs at least six. The four home associations hold, in practice, a veto: if all four oppose, nothing passes.

That detail explains much of VAR's character. This is a clause drafted inside a structure that demands broad consensus, not a product designed by engineers optimising for speed. Every phrase in it is the result of negotiation.

IFAB approved VAR in 2026. The system debuted at the 2026 Confederations Cup and entered the 2026 World Cup as an official refereeing aid. The Premier League only adopted it from the 2026/20 season. Which means that when England's top flight started using VAR, its refereeing teams already had three years of international data to draw on — and still made errors in exactly the kind of situation that international data had flagged.

VAR's trigger clause sits inside a single phrase: "a clear and obvious error, or a serious missed incident." Only four categories can be reviewed: goal or no goal, penalty or no penalty, direct red cards, and mistaken identity. The on-field decision stands unless the infringement meets the "clear and obvious" threshold.

That phrase is a trigger clause, like every other trigger clause in institutional life. Its value depends entirely on the mechanism that decides who has the authority to declare it satisfied.

Three tiers of verification for one incident

I work with a habit: never accept a single source, a single report, or a single frame. Every incident must pass through three tiers — provenance of the data, historical context, and deviation from the norm. The incident in the 34th minute on 30 September was a good test.

Tier one: provenance.

Four independent sources cover the incident: the audio published by PGMOL on 3 October 2026; the official match report; broadcast footage; and Liverpool's public statement. All four agree on the core event. So the event is not the point of dispute. The cause is.

The audio shows this: the VAR team processed the situation as though the on-field decision had been "goal," when the on-field decision was "no goal." The error lay in the labelling step. It did not lie in the camera or the calibration line. An input was mislabelled at the very first step, and every subsequent step of processing was correct — correct in a meaningless way.

I have seen this kind of error before. In 2026, as a first-year student working as a data-analysis assistant for FC United of Manchester, I found that the official statistics system had missed two penalty-area fouls in a match against Radcliffe Borough. I spent three days re-watching the full footage. My conclusion was not "the referee was poor." My conclusion was: the input sheet was missing a column. When a column is missing, the data still runs. It just runs wrong.

Tier two: historical context.

Before VAR, the same incident ended at the flag. The argument lasted two days and faded. What is new about the VAR era is not error — football has always had error. What is new is that error is recorded, replayed, and published with sound.

In 2026, after I named the wrong player as the recipient of a card in a university derby report, I spent six weeks logging 189 card incidents at the World Cup as reference data. The lesson was simple and deeply uncomfortable: refereeing error rates did not rise when VAR arrived; detected error rates did. The two figures are read as one another in almost every television debate.

Tier three: standard deviation.

This is the tier I trust least and check hardest. When a league publishes VAR's accuracy rate, the right question lies elsewhere: is that rate higher or lower than which benchmark, and where does the benchmark come from?

A 96% accuracy rate sounds excellent. But 96% of what? If the denominator is "situations in which VAR intervened," the measure assesses the quality of the occasions when the VAR team chose to act. If the denominator is "all situations that should have been reviewed," the measure assesses something entirely different — and usually much lower, because situations that were skipped never enter the denominator.

In Tottenham v Liverpool, the VAR team did not intervene incorrectly. The VAR team correctly declined to intervene in a situation they believed required no intervention. In any statistics table that counts only "number of interventions," this incident is almost invisible.

In 2026, I spent four weeks analysing Morocco's twelve matches at the World Cup and counted 87 tactical fouls. Their defensive system relied on cutting off players away from the ball rather than contesting directly, and their average card rate was 32% lower than that of European sides at the same stage — despite clearing the ball more often. Same logic: a team can commit few fouls in the report and still break an opponent's rhythm at every build-up. A low index is not automatically a sign of cleanliness.

By Euro 2026, I compiled 23 matches between 2026 and 2026 and found that Portugal received cards at a rate 41% higher in matches officiated by French referees. A UEFA refereeing researcher used that compilation when assessing the consistency of officiating teams. I mention it for one reason: if consistency between officiating teams is a measurable variable, then consistency of procedure must be measurable too. Right now, it is not measured.

What broke was the wording

The easiest conclusion after 30 September 2026 is: VAR is broken. That conclusion is convenient; it lets people rage at a machine instead of reading a text.

Consider the structure of the two words that caused the incident. In the VAR protocol, "check complete" means: the check is finished, there is no intervention. The problem is that the phrase does not specify which decision is being confirmed. It does not state the on-field decision. It does not state the VAR team's conclusion.

In a process where the two sides communicate only by intercom, under time pressure, in front of tens of thousands in the stadium and millions watching, ambiguous language is not a small detail. Ambiguous language is a design flaw. And a design flaw cannot be fixed by disciplining the operator.

This is where public debate goes off course. People punish the individual, then keep the procedure. People replace the VAR, and the incident recurs in another round, another league, another country.

I am not arguing for the defence of individuals. Darren England and Dan Cook were withdrawn from the following weekend's appointments, and that was correct under the regulations. But when an error recurs with different people in different places, the hypothesis of "poor individual" no longer holds. The hypothesis that remains is "poor design." That one is harder to accept, because it forces you to amend a document rather than a person.

There is a statistical trap I encounter constantly. Effort metrics in football — distance covered, sprints made — are packaged and presented as proof of professionalism. But running a lot is not automatically effective. A player can cover 12 kilometres in a match and not cut out a single pass. Attractive numbers are not necessarily true numbers.

VAR is the same. A league can publish a long list of correct interventions and look thoroughly modern. But if the trigger clause remains vague, that list measures diligence, not accuracy.

One contrast is worth remembering. In tennis, Hawk-Eye answers a binary question: the ball is in or out. There is no "clear and obvious" threshold, no referee's discretion. Football has no such luxury, because most VAR situations involve judgement rather than measurement. Precisely for that reason, football needs a tighter trigger clause, not a looser one.

Eight minutes in the same match

Curtis Jones was sent off in the 26th minute. The VAR team took every step correctly: they called the referee to the monitor, they pointed out the point of contact, they let the on-field referee make the final call — exactly in line with the principle that the referee on the pitch retains the decision. Nobody disputed the outcome, Liverpool included.

Which means that within eight minutes, the same VAR team operated one procedure correctly and one incorrectly. If VAR were a broken machine, it could not have managed the first. If VAR were a sound machine, it could not have botched the second.

Only one explanation holds: the tool is stable, the operators are competent, and the interface between the two is faulty.

I call that the gap between the law and its operators. My work begins precisely in that gap. A league is a system. Every refereeing decision is a variable. My job is simply the act of verification.

Three verifiable changes

I am not proposing that VAR be abolished. No data supports returning to an era when a flag was the final verdict, and the number of missed errors then was many times higher.

Three small changes below sit within IFAB's reach and can be measured within one season.

Require the VAR team to state the on-field decision aloud before starting the check. A single sentence: "The on-field decision is no goal." The cost is close to zero, and it eliminates the entire category of labelling errors.

Separate two kinds of error in the end-of-season report: judgement errors and procedure errors. Today they are lumped together, and because they are lumped together, only people are re-evaluated while procedures are not.

Publish the audio within 24 hours rather than four days. PGMOL did the right thing, only far too late — late enough for every misreading to sprout before the official recording appeared.

Semi-automated offside technology, used at the 2026 World Cup and introduced in the Premier League during the 2026/25 season, is an example of the right direction: it removes the manual labelling step from the process. But it addresses only one category. The other three remain in human hands — which means they remain in the hands of wording.

When data contradicts the eye, trust the data — but never forget to check where it came from. In the 34th minute of 30 September 2026, the eye was right and the data was wrong, simply because a label had been stuck in the wrong place. A card placed in the wrong position can change the current of an entire season. I have been the man who wrote that down incorrectly. I log every card, every minute of stoppage time, because a wrong figure repeated three times becomes a fact in the end-of-season report. VAR is not wrong. The people operating VAR are wrong. And that is exactly where my work begins.