Anfield, Mike Dean and 47 Decisions: When Data Enters the Dressing Room

The 73rd Minute at Anfield

In the 73rd minute of a 1-1 draw between Liverpool and Sunderland at Anfield in early February 2026, Sadio Mané received the ball on the right flank in a position that I — sitting in row seven of the press box, level with the edge of the penalty area — could see was clearly offside. The flag stayed down. The ball ended up in the net. The equaliser stood, and Anfield erupted in the way only Anfield erupts: a roar that was not pure joy but relief mixed with anger accumulated over weeks.

I did not roar with them. I opened my notebook.

Across 90 minutes plus stoppage time that night, I logged every decision made by referee Mike Dean: standing position, sightline, timing of the call, distance to the point of contact, direction of movement beforehand, and whether an earlier incident had primed the error. Forty-seven entries in total. Back at the hotel I built a spreadsheet by hand with 12 assessment criteria, then cross-checked it against television angles I had hastily recorded off the press-box monitor.

The result kept me awake: Mike Dean was correct on 46 of 47 decisions. A rate of 97.9 percent. And the single error — the offside at minute 73 — was the decision that settled the match.

That was the night I understood something that later became a professional principle: in football, average accuracy is a meaningless metric if you do not know where the error is distributed. A referee who is right 97.9 percent of the time but wrong in the 73rd minute of a 1-1 draw will be remembered longer than a referee who is right 88 percent of the time but wrong in the 12th minute of a 4-0 win. The right question is not "is this referee good or bad" but "what type of incident does his error fall into, and does that type of incident regularly decide results".

From that night, the way I wrote about referees changed entirely. I stopped describing a decision. I started describing a decision-making system, and pointing at where that system fractures.

Fifty Years of Accumulated Argument Before the VAR Room Existed

To understand an offside at Anfield in 2026, you have to place it in a much longer current than the match itself.

In 2026, at Wembley, Geoff Hurst's disputed goal in extra time of the World Cup final between England and West Germany opened a debate that ran for half a century: had the ball crossed the line. No tool could answer it. In 2026, in Mexico City, Diego Maradona scored with his hand against England and called it the "Hand of God" — a phrase both self-mocking and defiant, because he knew no device could contradict one man's testimony.

But it took until 27 June 2026, in Bloemfontein, for football to reach the limit of its own tolerance. Frank Lampard struck the ball, it hit the crossbar and bounced clearly over the goal line against Germany. The referee did not award the goal. England were eliminated. The whole world saw the truth on television; only the man with the authority to decide did not.

That was the moment the conservative argument collapsed. You cannot keep saying that error is part of football's romance when the error could be corrected by a camera placed at the right angle.

Goal-line technology was approved and introduced from the 2026-13 season, appearing in the Premier League and then at the 2026 World Cup in Brazil. But goal-line technology solved only a very small part of the problem: it answered "did the ball cross the line", not "was there an offside", not "was there a foul in the box", not "was the red card correct".

The International Football Association Board (IFAB) approved VAR trials in March 2026. The first experiments took place in lower-pressure competitions and expanded gradually. By the 2026 Confederations Cup, VAR appeared at a senior international tournament. And at the 2026 World Cup in Russia, it stepped onto the biggest stage of all.

What matters is the intervention threshold. VAR was never designed to be right in every situation. It was designed to correct a "clear and obvious error" — in a goal, a penalty, a direct red card, or a case of mistaken identity. That is a philosophical choice, not a technical one. It accepts that small errors will be left standing, in exchange for a match not being shredded into hundreds of replays.

And that philosophical choice is where every subsequent argument was born.

Twelve Criteria and the Line Between Technical Error and Perceptual Error

Back to my spreadsheet from that night in early February 2026.

I split the 47 decisions into four groups: geometric decisions (offside, ball out of play, throw-in direction), contact decisions (fouls, degree of contact), disciplinary decisions (yellow cards, red cards, warnings), and game-management decisions (stopping play, playing advantage, handling confrontations).

Then I applied 12 criteria to each entry: the referee's body position, whether his sightline was blocked by a player, distance to the point of contact, speed of the incident, number of players in the contest zone, reaction time measured in fractions of a second, direction of the ball, actual degree of contact, the incident history within the match, the impact on the result, consistency with earlier decisions, and finally the level of agreement within the officiating team.

The result showed something I have reused in many analyses since: most refereeing errors are not technical errors but perceptual errors.

A technical error is when the referee stands in the wrong place, is blocked, or lacks a sightline. That kind can be fixed with physical training, movement patterns, technology. Perceptual error is entirely different: the referee sees correctly, but the brain misinterprets. That is the anchoring effect — when an earlier decision shapes how you read the next incident. That is confirmation bias — when you have already decided that Team A is playing rough, you read Team A's challenges as more serious than they are.

Among Mike Dean's 47 decisions that night, I counted nine incidents showing signs of perceptual error. He got eight of the nine right. The ninth was the offside in the 73rd minute.

This matters because it completely changes how you evaluate a referee. If you only count the correct rate, you will conclude Mike Dean is one of the Premier League's best referees — and you would be right. But if you count the distribution of errors by impact, you see a different problem: his errors cluster in geometric decisions in dangerous areas, where a single miss is enough to change the result.

A camera finds the error, but a human being finds the cause. A system that only counts errors will always fail to improve quality, because it cannot distinguish an error caused by missing information from an error caused by misprocessing information.

101 Seconds: The Real Cost of a Review

In June 2026, I accepted an assignment to analyse VAR for World Cup matches in Russia.

On 16 June 2026, in the France versus Australia match, Antoine Griezmann went down in the penalty area. The referee initially waved play on. VAR intervened. The penalty was awarded. The stands booed, and the commentators around me began the familiar refrain: VAR is killing the rhythm of the game.

I did not argue with them. I started a stopwatch.

I measured the time from the moment the referee put his hand to his ear to the moment he pointed at the spot. The average across reviews at that tournament: 101 seconds. Then I did something nobody in that press room did that day: I compared total stoppage time in matches with VAR against matches without it in comparable competitions.

The result: average stoppage time increased by only 2 minutes 37 seconds.

Two minutes thirty-seven. That is the price of correcting errors.

I wrote that piece with a simple argument: if you trade 2 minutes 37 seconds for a reduction in wrong goals, wrong penalties and wrong red cards, you are getting a bargain. VAR's problem was never time. VAR's problem is the feeling of being disenfranchised — the moment when a full stadium must fall silent and wait for a verdict from someone 300 kilometres away.

I was once a VAR sceptic, and that is why I understand those who hate it. VAR sceptics do not object to accuracy. They object to the interruption of emotion. You are screaming with joy, and suddenly a line is drawn on the big screen, and your joy hangs suspended for 101 seconds.

I once spent four months delaying the publication of a study because I did not trust my own numbers. That period taught me that perfectionism has two faces: it protects you from error, and it also costs you the moment. I learned to set an internal deadline two weeks before the publication deadline, and to send drafts to two trusted colleagues for early challenge.

But there is one thing I never conceded: the process had to be verified before publication, even when the whole world was waiting.

89 Matches Without Crowds

In June 2026, football returned after lockdown. The stadiums had no spectators.

I agreed to contribute analysis to an independent study on the effect of crowds on refereeing decisions. We collected data from 89 Premier League matches before and after the pandemic, controlling for team quality, fixture density and squad availability.

The result: yellow cards fell 23 percent. Penalties rose 31 percent.

Those two numbers move in opposite directions, and that is the most interesting part.

The first explanation, the most accessible: without crowd noise, referees no longer face social pressure from 40,000 people. In football, a collective roar is a form of false evidence — it does not prove a foul, but it creates an environment in which not giving the foul feels like a betrayal. When the roar disappears, referees book fewer players.

The second explanation: without the roar, referees also lose a confirming signal. In penalty-area contacts, crowd reaction is a form of supplementary information. Without it, referees fall back on their own sightline — and in many contacts where a foul genuinely occurred, the reaction of players plus the noise helped them decide faster.

An empty stadium does not lose its soul; it only returns the soul to its rightful owner. Without spectators, the match no longer belongs to us. It belongs to the players, to the laws, and to the people who must decide with nobody behind them to hide behind.

I initially intended to keep that research private. Perfectionism made me delay publication for four months, repeatedly re-running the numbers, questioning whether 89 matches were enough, whether the control variables were sound. When it was finally published, it was cited by UEFA's data analysts.

But those four months were four months in which my readers had nothing to read. And that was a process-management failure, not a data failure.

78 Touches and the Method of Micro-Observation

On 21 November 2026, in Qatar, I sat in the VIP stand for England versus Iran. The score was 6-2. Bukayo Saka scored a hat-trick, and every bulletin that night revolved around him.

I spent the whole match watching someone else.

Jude Bellingham, nineteen years old. In my notebook I logged 78 touches. But the more important number: 41 of those 78 touches were one-touch. He almost never held the ball longer than three seconds.

That is a decision-making structure, not a statistic. A one-touch reception means the player has already read the situation before the ball arrives. Holding the ball for under three seconds means he does not need time to process, because he processed before receiving. At elite level, the gap between a good midfielder and an outstanding one is not technical ball control. It is the number of seconds that player takes away from the opponent.

That night I called a former scout I trusted to confirm what I was seeing. After the tournament I wrote a long-form series on Bellingham, predicting he would become the best central midfielder of his generation — before most major European newspapers said so.

I tell this story in a piece about refereeing because it is the same method. Micro-observation. Not looking at the final outcome, but at the structure beneath the outcome. The way I logged Mike Dean's 47 decisions in 2026 is the way I logged Bellingham's 78 touches in 2026: the same professional habit, the same belief that truth lives in frequency, not in the moment that gets replayed.

A referee's power does not come from the whistle, but from the ability to read the situation. And the ability to read situations, in anyone, is built from small observations accumulated over years.

Emotion Is Not the Enemy of Law

This is the part where I have to argue against myself.

For years I was seen as an advocate for technology. The reality was more complicated.

After the 2026 World Cup, I began receiving critical emails from supporters, mostly from England. Their content was not "VAR is wrong". Their content was "I no longer know when I am allowed to celebrate".

At first I dismissed it. That is emotion, not argument. But supporter emotion is also data, just data that has not been encoded.

I started asking VAR opponents about the specific moment that drove them away from football. Almost all of them described the same scene: the ball goes in, they jump up, then they see the referee put his hand to his ear. The joy freezes. And even when the final decision is a valid goal, the joy does not return. It becomes a kind of administrative relief.

This is the biggest blind spot of the data-analysis community, myself included for many years: we can measure accuracy, but we do not measure the cost of waiting.

If you trade 2 minutes 37 seconds for accuracy, you must understand that those 2 minutes 37 seconds are not evenly distributed. They cluster in the most important moments — goals, penalties, late decisions. You are not selling scattered time. You are selling time at the peak of emotion.

That is why I oppose proposals to extend VAR to corners and throw-ins. Every expansion of the intervention scope increases the emotional cost multiplicatively, while the accuracy benefit rises only linearly. At some point, you are paying far too much for very small gains.

And here is what I had to correct in myself: in my first piece on VAR, I made a reasoning error. I compared review time with stoppage time as though the two were equivalent in experience. They are not. Stoppage time is an addition the viewer can anticipate. A review is a subtraction the viewer cannot anticipate. An independent commentator pointed out the flaw, and I wrote a correction.

I was wrong once, and I said so. Not to appear humble, but because in this work, the person who never admits error is the person who has stopped observing.

The Whistle Does Not Create Power

After seven years of note-taking, one thing has become clear to me.

Football does not have a refereeing problem. Football has a problem with how it evaluates referees.

We evaluate referees through the memory of viewers, and viewer memory is a poor instrument: it records the mistake but not the context, it remembers the wrong call in the 73rd minute but not the 46 correct decisions before it, and it cannot distinguish between a weak referee and a good referee facing a hard situation.

If we accept measuring xG to evaluate chance quality, why do we not accept measuring "probability of a correct decision" by incident type to evaluate referees? Not to replace humans, but to direct training resources to the right place. A referee with a low correct rate in penalty-area contacts should be trained in that exact area, not punished by demotion.

This is the largest untapped investment in modern football: the game spends hundreds of millions of pounds on player data, hires entire analytics departments to optimise a corner, yet still uses a manual marking sheet to assess the decision-maker with the greatest influence on the result once the ball is rolling.

And this is what I want to leave behind.

A referee's whistle is not the source of power. The power lies in the fact that at that moment, nobody in the stadium is allowed to object. There is no other person among the 22 on the pitch with the right to say "again" and have the whole match change accordingly.

If you give anyone that kind of authority without giving them a serious feedback system, you are not protecting the laws. You are protecting an asymmetry.

An empty stadium does not lose its soul; it only returns the soul to its rightful owner. And a referee with no crowd roaring behind him must still make the right call. Our study of 89 matches proved it: remove the pressure, remove the bias, and decision quality shifts in ways an entire football industry had failed to anticipate.

If that is true when the stands are empty, then the next question is this: how much of what we call a referee's "instinct" was only crowd noise, internalised?

I do not answer that question here. I leave it to those who will keep taking notes in row seven.