The Law Is Not in Memory: 523 VAR Matches, the Armpit Zone and the Price of One Wrong Sentence at the 2026 World Cup

On 16 June 2026, in the 55th minute at Kazan Arena, referee Andrés Cunha of Uruguay walked out of the penalty area and headed for the monitor beside the touchline. In a small studio in Valencia, I was speaking into a microphone, saying that Josh Risdon had only let the ball touch his shoulder, that a shoulder is not a hand, that it was a clean challenge and there was no penalty. I said it in the voice of a man who had been reading the laws of football for thirty years. I was wrong.

Four million listeners heard me be wrong. The editorial desk had to read a correction within twelve minutes. The colleague sitting next to me, twenty years my junior, said something I still keep in my notebook: "The armpit zone has counted for a long time now." I opened my laptop during the break. He was right.

It was the first time in thirty years that I had been contradicted live on air. What I lost was not an answer. What I lost was credibility — something I had built over thousands of hours of reviewing footage, and dropped with one unverified sentence.

The 2026 World Cup was the first World Cup to use VAR. IFAB, the body that governs the laws of the game, formally approved the technology on 3 March 2026 at its meeting in Zurich, after years of trials in domestic leagues. The France–Australia match in Group C on 16 June 2026 in Kazan became the first match in World Cup history in which a penalty was awarded after the referee reviewed video footage.

Earlier, in the 55th minute, the ball from a French cross struck Josh Risdon inside the penalty area. Cunha did not give it at first. He received a signal from the VAR team, walked to the monitor, watched two angles, and pointed to the spot. Antoine Griezmann converted in the 58th minute. Mile Jedinak equalised from another penalty in the 62nd, also after VAR intervention, this time for a handball by Samuel Umtiti. Paul Pogba scored the 2–1 winner in the 81st minute, and that goal was confirmed by goal-line technology — a technology IFAB approved in 2026 and which first appeared at a World Cup in 2026.

I retell this sequence not to tell a story. I retell it to point at something very few people in the commentary trade are willing to admit: the biggest errors in VAR controversies do not sit with referees. They sit with the people outside the pitch who explain the law without checking the law. The referee at least has a monitor, a VAR team, a report. The commentator has only memory — and memory is the worst of all data sources.

At the 2026 World Cup, 29 penalties were awarded, the highest number ever recorded at a single World Cup up to that point. That figure is routinely cited in the media as proof that "VAR ruined football". But when I sat down with the raw data — every penalty, every minute, every reason — the picture looked very different. Most penalties came from incidents the on-field referee had missed because his view was blocked, not from incidents he had seen and judged incorrectly. Those are two entirely different categories of error, and merging them is a serious analytical mistake.

To talk about handball, you must first talk about the boundary. From the mid-1990s, when I learned the laws to practise as an expert, my mental model was simple: the arm ran from the shoulder down, and the shoulder was part of the torso. I kept that model for more than twenty years without anyone challenging it, because most incidents fell into clearly contested zones — either the arm or the chest. The armpit zone almost never appeared in the handball debates of the 1990s and 2000s.

In the 2026/20 edition of the Laws of the Game, IFAB set out a clear definition in Law 12: the upper boundary of the arm is in line with the bottom of the armpit. In other words, the armpit zone belongs to the arm, not the shoulder. A ball touching the armpit zone is a ball touching the hand. Later amendments tightened it further: an arm extended above the shoulder is treated as making the body unnaturally bigger, and in the edition effective from the 2026/22 season, a player who scores or creates a goal with a handball — even an accidental one — is still penalised.

What I want to stress here is not the content of each amendment. What I want to stress is the structure of the change. Over thirty years, the boundary between arm and shoulder has shifted at least three times, in different directions. A practitioner who keeps his mental model of the law frozen across three shifts is no longer an expert. He is an outdated library.

And I was that outdated library, for twelve minutes, in front of four million people.

After the night in Kazan, I did something I should have done ten years earlier. I built a law-reference table updated year by year: each row listing the article number, the content, the date of publication, the date of effect, and a note on the previous version. That table now runs past two hundred rows. Every time I write or go on air, I open that table first, not my memory.

The law is not in memory. It is in data.

But a law table only solves half the problem. The other half is a question no table can answer: applied to two different matches, does the same article produce the same decision? That is why I started my own data project in August 2026.

I logged every VAR-involved decision in La Liga and the Champions League: error code, minute, score at the time, distance between the referee and the incident, ball speed at contact, and the time from VAR intervention to the referee's final decision. I did this alone, roughly twelve to fifteen hours a week, with one Excel file and two monitors.

By March 2026, when the pandemic stopped European football, I had 523 matches in the database. That number was large enough to start talking about patterns rather than anecdotes.

One match is a story. Five hundred matches are a law.

The first finding that forced me to rewrite how I commentate: within the group of overturned offside-related decisions, 74 per cent of cases had an average response delay of 47 seconds from the ball entering the net to the referee formally changing the decision. That 74 per cent does not say referees are wrong. It says the system communicates badly.

Picture a spectator in the stand. The ball hits the net. He screams. Three seconds later he sees the referee put a hand to his ear. Fifteen seconds later he sees the referee draw a rectangle in the air. Forty-seven seconds later he finally learns whether the goal stands. During those forty-seven seconds he is not neutral. He is building a story — and the story he builds depends on which team he supports.

This is the point most VAR analysis skips. VAR's problem is not accuracy. Its problem is latency, and the information vacuum that latency creates. During those forty-seven seconds the crowd is not waiting for data. The crowd is waiting for a reason to believe it is being cheated.

I published a 48-page report on my personal blog in April 2026, with the Excel file attached so readers could check it themselves. In that report I proposed a hard limit: no video review should exceed 30 seconds; if no clear conclusion emerges within 30 seconds, the on-field decision stands. The proposal was not adopted. But the Valencian Football Federation invited me to advise a regional working group on refereeing process reform, and I accepted.

I accepted for a very specific reason. If a time limit will not be adopted, then at least young referees must be trained to decide faster under pressure — and to train that, you first need data on where they are slow. That is the part I can contribute.

Based on my experience watching matches in La Liga during the 2026/20 season, I noticed another pattern that my data was not strong enough to conclude on, but strong enough to raise as a question. Controversial decisions in matches involving the biggest clubs tended to have longer review times, and the rate of overturning the original decision in favour of the bigger club ran slightly above the sample average. I say "slightly" deliberately, because the margin of error in the sample does not allow me to say more.

What I can say more strongly is this: crowd pressure and media pressure are real variables, not a conspiracy theory. A referee working in front of 80,000 people in a large stadium, knowing his decision will be replayed twenty times on national television, processes information differently from a referee working in a 15,000-seat ground. Nobody orders him. No call comes from above. There is only the weight of consequence, and that weight is not evenly distributed.

This is why I never use the word "fixing" when discussing decision disparities. Fixing is a criminal accusation that requires criminal evidence. A decision disparity is a psychological and social phenomenon that requires data, not accusation.

Referees do not need to be protected. They need to be understood through correct numbers.

Around the same time as my data project, I noticed a similar category of error unfolding in a completely different subject: how we talk about players returning from injury.

In a major tournament cycle, the media has an almost mandatory template: a returning player must "prove himself", must "handle the pressure", must "carry the team". It sounds like praise for mental strength. But looked at through data, it is an unreasonable demand.

A player returning from an ACL injury or a long-term muscle injury does not come back with the same body. Sports medicine research published in specialist journals through the 2010s and early 2020s shows re-injury rates significantly higher in the group of players who return too early compared with the group given full rehabilitation time. I am not citing a specific figure here because the studies define "too early" differently, and blending those definitions is precisely the error I am criticising.

But the principle is clear. When a player returns in the 70th minute of a big match and is pressed relentlessly for the final twenty minutes, the question is not "does he have character". The question is: who decided to send him on in the 70th minute, based on what data, and did that data account for how many days he had been training at full intensity.

I have watched hundreds of matches and logged these situations. The pattern repeats: returning players tend to be introduced at the stage the team needs most, not the stage the body is most ready for. That is a rational tactical decision and a poor medical one. Both can be true at once.

And here is the link to the armpit story. In both cases, people are working from an old model — a model of the law, or a model of the human body — and failing to update it with new data. The only difference is this: when I was wrong about the armpit, I lost credibility. When a club is wrong about a player's return date, the player loses part of his career.

In both cases, the decision-maker does not pay the price.

Now I want to address the part I consider most important, and also the part that took me longest to accept.

After I published the 523-match report, I received a great deal of mail. Most of it was technical feedback, some was criticism, some was invitations to collaborate. But there was one group of letters I read again and again. They came from serving referees. Their content was almost identical: they did not dispute my data. They disputed the fact that I published data without stating the context of each match.

One referee wrote to me: "You counted the number of times I took longer than average to review footage. Did you count the number of times I was ill that week? Did you count the number of times I had to officiate a match where one team had mentally collapsed from the 20th minute?"

I had no answer to that. And that is when I realised I was repeating my old mistake at a different level. I had abandoned memory for data. But data without context is just another form of memory — the memory of a spreadsheet.

I once wrote a piece defending referees using bad statistics. That was the moment I betrayed my own principle.

I corrected the report. I added a new chapter, eleven pages long, listing everything my data cannot measure: referee health, pitch quality, pre-match pressure, congested schedules, and most importantly — the fact that I only logged decisions where VAR intervened, meaning I only saw the tip of the iceberg. Every correct decision nobody mentioned is absent from my database.

This is a textbook selection bias, and I carried it for ten months without noticing.

After the correction, I drew a conclusion I think is useful for anyone reading refereeing analysis, including my own: always ask how the sample was selected. If a piece says "referee X has the highest overturn rate in the league", the first question must be: across how many matches, over what period, and were the decisions that were not overturned counted.

One match is a story. But a biased dataset is also a story — and a more dangerous one, because it wears the clothing of numbers.

There is one more dimension I want to raise, even if it may irritate some colleagues.

The football commentary industry has built a professional model based on rapid reaction. Within fifteen seconds of a controversial incident, the commentator must say something. Silence is read as a lack of expertise. Hesitation is read as a lack of confidence. So the commentator speaks — drawing on the only things available in those fifteen seconds: memory and instinct.

I lived inside that model for thirty years. And I believe that model is generating a systematic layer of distortion in public understanding of the laws.

Think about the audience of a major match. Tens of millions. Every one of them hears at least one commentator explain a legal situation during the game. If the error rate of that commentary layer is 10 per cent — a figure I consider optimistic — then in a single major match, millions of people absorb incorrect information about the law and carry it into the next match.

A referee's mistake affects one match. A commentator's mistake affects millions of people for years.

That is why I propose a professional standard I call "citing the law". Specifically: whenever a commentator or expert asserts that an incident is right or wrong under the laws, that person must be able to name the article number and the edition of the laws in force. If they cannot, they must say clearly: "I need to check that."

It sounds rigid. But remember that we accept a similar standard in other fields. When a doctor discusses a drug, we ask for the name and the dose. When a lawyer discusses a case, we ask which statute. Only in football do we accept "as I recall…" as a valid argument.

And this is the counterintuitive part of the whole story.

The popular reaction after VAR controversies is to demand more technology: sensors in the ball, higher-speed cameras, artificial intelligence that determines offside automatically. I am not against technology. But I believe that reaction misdiagnoses the disease.

Modern football's problem is not a lack of data. The problem is that data exists but does not reach the person who needs it, at the moment they need it. An automated offside system with 99 per cent accuracy will still generate controversy if the stadium must wait 47 seconds to learn the outcome and nobody explains what is happening.

Put another way: VAR's problem is not the accuracy of the decision, but the speed and quality with which that decision is communicated.

And if that is right, then the cheapest, fastest, most effective solution is not buying more cameras. The solution is changing how the decision is communicated: display the reason on the big screen the moment the referee begins a review, announce the outcome over the stadium PA, and impose a hard limit on review time.

Those three changes need no new technology. They need administrative decisions. And they could be implemented from next season.

At 67, I no longer have enough time to wait for a technological revolution. I do have enough time to re-read an article before speaking about it.

At 67, I do not need to remember everything. I need to know how to find what is correct.

I once got one sentence wrong and lost an entire reputation. If only I had known this back then.

There is one small detail from the Kazan story I have never told publicly.

After the editorial desk read the correction, I stayed in the studio for another forty minutes. I reopened the clip of that incident and watched it twenty-three times. On the eighteenth viewing, I noticed what I had missed from the very beginning: the question was not where the ball touched Risdon's body. The question was what position his arm was in relative to his body, and whether that position made his body unnaturally bigger.

That is an entirely different question. And it was a question I never asked, because I was busy answering a different one — the question about the boundary between shoulder and arm.

This is the biggest lesson I took from that night, and it has nothing to do with looking things up. When a controversial incident appears, people tend to pick the easiest question to answer and then answer it with total certainty. The question about the shoulder–arm boundary is easy to answer. The question about arm position is far harder, because it depends on movement, on ball speed, on whether the player was turning.

And in football, as in every other field, the easy question is usually preferred to the right one.

I think this is the trap that VAR debates in Vietnam and in Spain are both falling into. People argue endlessly about one easily measurable dimension — where the ball made contact — and say very little about the hard-to-measure dimension, which is the one that actually decides the case under the laws.

Since that night, I have set a rule for myself: before answering any question about the laws, I must ask myself whether the question I am answering is the most important question in that incident. If it is not, I must tell the person asking that their question is not yet the right one.

That rule has cost me some invitations to appear on air. But it has saved me from ever having to read another correction.

I want to close with something in the nature of a proposal, aimed at the people working in football in Vietnam — where I follow matches through published reports and statistical data.

Vietnamese football is at a stage where refereeing decisions receive unprecedented attention, especially in major matches and in fixtures involving the national teams. This is a good moment to build a different habit, before the bad habits set in.

Specifically, I propose three things that can be done immediately without expensive technology. First, every report on a controversial incident should carry the article number of the law being applied and the edition of the laws — this forces the writer to check, and forces the reader to learn how to verify. Second, broadcasters should maintain a law-reference table updated each season, like the table I built after 2026, so commentators do not have to rely on memory while on air. Third, competitions should publish data on review times for every match, because that data forces all parties to discuss the real problem instead of arguing about feelings.

None of those three requires next-generation VAR. All of them require a decision that looking things up is part of the job, not a sign of weakness.

It took me thirty years to learn that. I hope the next generation of referees and commentators will not need to take that long.

A shocking decision is not reckless if it is built on five hundred foundations.

And a wrong sentence does not need five hundred foundations to collapse. It only needs one moment of overconfidence.