VAR Notebook: From Kazan 2026 to the V.League, How a Refereeing Decision Is Actually Manufactured

On 27 June 2026, at the Kazan Arena, in the second minute of stoppage time, Kim Young-gwon stabbed the ball into Germany's net. The assistant referee raised his flag. I was seventeen, watching from Seoul, and I did not scream like my classmates did. I stared at the background of the frame: referee Milorad Mazic standing still, hand to his earpiece, waiting on the VAR room. Inside Kazan, the roar burst and then choked — tens of thousands of people who did not know whether they were allowed to celebrate. Forty seconds later the goal stood. South Korea beat Germany 2-0 and the defending champions went out. That night I downloaded all 64 matches and started a notebook that recorded nothing but VAR. Two weeks later it ran to 47 pages: minute, score, referee position, number of reviews, who decided last. I did not know what I was doing. I only knew the thing that had just changed at Kazan was not in Kim Young-gwon's feet. To understand Kazan you have to read Law 11 again. By 2026 IFAB had tightened the offside definition: a player in an offside position is penalised only if he interferes — touching the ball, impeding an opponent, or gaining an advantage. Assistants were told to hold their flag when the situation was unclear, deferring to the VAR room. At Kazan, the decisive detail in the 90th minute was that Toni Kroos got the final touch before the ball reached Kim. One touch by a German midfielder pulled a Korean striker out of an offside position. Nobody in the stadium saw it with the naked eye. One camera behind the goal saw it. The VAR protocol FIFA applied in Russia covered four categories: goals, penalty kicks, direct red cards, and mistaken identity. The intervention threshold was a clear and obvious error, or a serious missed incident. The referee still made the final call; VAR only recommends. On paper this preserves the referee's authority. In practice it redistributes authority to people sitting in a closed room — people who control which camera angle is chosen and when. Russia 2026 was the first finals with VAR. FIFA reported 20 decisions changed after intervention, and the tournament ended with more penalties than any previous edition. Four years later FIFA added semi-automated offside, cutting verification from tens of seconds to a few. Alongside that came a less noticed change: new guidance on calculating stoppage time produced nearly 27 added minutes in England versus Iran on 21 November 2026. In the same window, the Netherlands-Argentina quarter-final on 9 December 2026 in Lusail, refereed by Antonio Mateu Lahoz, produced 18 yellow cards and a red for Denzel Dumfries — the highest card count in a World Cup knockout since Portugal-Netherlands in 2026, which had 16 yellows and four reds. Those three data points do not sit together by accident. They show football changing three things at once: how a foul is identified, how time is counted, and how human emotion on the pitch is managed. Each has a price, and the price is usually paid by the audience. In 2026, with the pandemic freezing every league, I was nineteen and had just started a broadcasting degree. With no football to watch, I reopened the Kazan notebook and turned it into a coding project: 1,247 refereeing decisions from the 2026 World Cup and three K League 1 seasons, each row carrying fourteen fields. It took four months. I produced a 30-page analysis. The finding that kept me at my desk longest I called the compensation effect. In my dataset, after a team suffered a wrong decision, the rate at which they received a soft penalty within their next two matches was 89 percent. The number was too clean to trust. I checked it three times and still got 89 percent. I never dared publish it, afraid a test was missing. Only when I worked at Qatar 2026 as a data assistant did I understand why I had been statistically right and humanly wrong. The 2026 season had no crowds, but it had one very large ear. With the stands empty, every other sound became data: the referee's breathing after a sprint, a player swearing, the ball against a post, and the long silence before a decision is announced. I learned that a refereeing decision has two lives: one on the pitch, lasting seconds, and one in the technical room, lasting minutes. Only the second is recorded as data. A referee does not confirm what is true; a referee confirms what three machines — the camera chosen by the broadcast director, the software in the VAR room, and his own body — permit him to confirm. A decision is not made on the pitch. It is made in three separate places and converges on one person, in less time than an away end takes to breathe. At Qatar I sat in the media tribune of Al Bayt for the opening match. The screen in front of me carried one feed. The screen at home carried another. Not because of technology, but because of the director. A World Cup match has dozens of cameras, a group of them dedicated to VAR — fixed, calibrated, uncut. Television audiences almost never see those in real time. They see the frame the director chose, at the moment the director chose, at the speed the director chose. I do not watch the goal; I watch the camera angle watching the goal. That habit began at Kazan and became method. When a decision is controversial, my first question is not whether the referee was right, but which angle the audience was shown before they concluded. The same contact looks different from behind the goal and from the touchline. One touch, two conclusions, both broadcast within thirty seconds. That is why I began building files on individual referees — not to attack them, but to predict them. In the summer of 2026, aged twenty, I wrote a 2,100-word piece on Raheem Sterling's penalty in the England-Denmark semi-final at Wembley on 7 July 2026. Referee Danny Makkelie awarded a penalty in the 104th minute after contact between Sterling and Joakim Maehle; Harry Kane's kick was saved by Kasper Schmeichel and he scored the rebound. I pulled Makkelie's previous five matches and found a pattern: four penalties awarded for contact in the box, all following one principle — the ball was live at the moment of contact. For Makkelie the question was never whether there was contact, but whether the ball was dead. Live ball, minimal contact, still a penalty. The piece ran in a university journal, a Seoul sport editor shared it, and within a week I had a part-time commission from a football site. That was when I understood data does not have to be perfectly correct to be useful. It has to be consistent. To price VAR I chose seconds rather than cards. A review does not just cost time; it costs rhythm. Football is a game of flow, and a broken flow forces players to rebuild their mental state. In Qatar, second-accurate stoppage time pushed match length to unprecedented levels. England-Iran ran nearly 27 added minutes, a figure that ten years ago would have been treated as a faulty clock. Behind the length sits an economic question. A 30-second advertising slot in the first half is priced differently from one at 90+8. When matches stretch, the value of each minute shifts and the gains are unevenly distributed. Broadcasters sell extra inventory. Organisers gain extra debate. Players pay with muscle. Referees pay with credibility, because their decision sits at the end of a timeline they do not control. The referee is the fastest reader of the match; I am only one beat slower. But being one beat slower is an advantage, because I am allowed to rewatch. I rewatched Sterling's penalty eleven times from four angles and realised the contact looked heavier on the main camera than on the reverse. No camera lies. Only the person choosing the camera has intent. At this point the story leaves the World Cup and enters Vietnam's V.League 1. From the 2026 season VAR formally arrived in the national top flight after years of infrastructure preparation and referee training. That is genuine progress and should be stated plainly. But the progress is constrained by three variables no league solves by buying hardware. The first is camera count. A match checked with four or five angles produces different outcomes from one with twelve. The clear and obvious threshold is not a physical constant; it depends on how much evidence you hold to overturn a call. With fewer angles the threshold rises automatically, meaning real errors survive and false ones are created by the absence of information. The second is public data. When I coded 1,247 decisions in K League 1, I did it by hand from match reports and video, because no open database aggregates referee decisions round by round. A league that does not publish decision data cannot accumulate knowledge about itself. Every season the same arguments return with new names attached. The third is audio. The Premier League began releasing VAR communication through its Match Officials: Mic'd Up programme from the 2026-24 season after years of pressure. UEFA introduced a captain-only rule for approaching referees at Euro 2026, a trial IFAB later approved internationally. Neither change makes referees more skilled. Both change incentives: when a recording will be published, referees speak shorter, clearer, and with fewer excuses. No paper report achieves that. The V.League does not lack cameras; it lacks a data publication protocol. That is the systemic difference between a league learning VAR and one that has passed that stage. The missing piece is not technology but the habit of turning a decision into a searchable, challengeable record that can be used to train the referees themselves. Data never commits a foul; the writer is the one who gets carded. If my 89 percent is wrong, the culprit is not the spreadsheet but the person who defined what a wrong decision is. Which brings me back to confront my own model. At Qatar 2026 I used the compensation effect to predict referees would limit cards in the group stage to protect match flow. It held in 26 of 36 group matches. Then the Netherlands-Argentina quarter-final collapsed it entirely with 18 yellows and a red in 120 minutes. No spreadsheet of mine predicted that, because I had not entered the one variable that mattered: the accumulated tension of a knockout both teams treated as their last. In December 2026, after the tournament, I spent three weeks interviewing two former FIFA referees. Both said the same thing in different ways. Referees are not swayed by crowds in the sense of fearing them. They are swayed by memory of the match in progress — by a sense of when it slipped beyond control. One showed me how to read body language: the referee starts walking more between incidents, starts talking to players with his hands rather than his voice, and starts looking toward the benches rather than the ball. Those three signs appear fifteen to twenty minutes before a match boils over. I had a predictive indicator I had never coded. The more I work, the more I suspect my compensation model may be a statistical illusion. The problem is definition. I called a decision wrong when it contradicted my conclusion after rewatching, but I never had access to every angle the referee had. If part of those 1,247 rows were not actually wrong, the denominator of 89 percent was distorted from the start. A serious study needs a control group: matches where that team suffered no wrong decision, to see their baseline rate of soft penalties. I have not done that part. I owe it. Discipline is not punishment; discipline is a way of reading a match. After rewatching hundreds of incidents, I noticed a yellow card is rarely used to punish behaviour that has happened. It is usually used to shape behaviour in the next forty minutes. A card in the 20th minute writes an implicit contract with both teams: this is my limit, adjust to it. A card in the 88th minute liquidates a contract that has failed. The same card, two entirely different meanings. What VAR cannot fix is that structure. Technology intervenes in judgement errors; it does not intervene in match management. The paradox is that more technology means the referee must manage more things. He manages a match, a technical room, a crowd rewatching on phones, and a broadcast director deciding what goes to air. Four parties, one man, ninety minutes. It should also be said plainly that the romantic argument that humans are entitled to error in modern football is mostly a convenient argument for organisations that do not want to publish data. When broadcast rights are priced per match and sponsorship deals are signed on a club's airtime, preserving an unpublished error is no longer respect for human frailty. It is an economic choice dressed as a cultural value. Kazan cancelled a goal, but it opened an eye. If I had to compress a decade of watching combat sport and football through a refereeing lens into one line, that is the line. Every controversy leaves behind a fragment of data. The problem is that almost nobody collects it, and so we argue again from scratch the following season. There are variables I never entered into the spreadsheet. The pressure of a host nation eliminated early; heat and humidity shifting the threshold at which contact counts as a foul; the acoustics of a closed roof; the career stage of the referee; and league culture — a referee working where every round is a relegation decider behaves differently from one working where the table settled in March. I list these not to excuse the model but to state its limits. A model that declares its limits is more useful than one pretending to have none. Back to the V.League. The thing worth doing before buying more cameras is not buying more cameras. It is establishing a public decision record for every match: the incidents checked, the incidents intervened on, the incidents upheld, and the reasoning written in the language of the laws. Audio need not be published immediately. A catalogue is enough. Once that catalogue exists across three seasons, the league will own something it currently lacks and cannot buy: a baseline for knowing whether its referees are improving or regressing. And the question I want to leave is not whether VAR is right or wrong. It is this: if every decision became searchable, who loses most when the data goes public — the referees, the organisers, or us, the people who grew used to arguing without evidence? The law is the only thing that never enters stoppage time. Everything else does. Including the audience's belief.