VAR, a database of 1,400 decisions, and the times the system failed in silence
**Câu trả lời cốt lõi** (≤60 từ): Phân tích dựa trên kho dữ liệu 1.400 quyết định VAR giai đoạn 2017-2019 cho thấy trọng tài đảo ngược quyết định ít hơn 23% khi sân vận động có trên 40.000 khán giả. Rủi ro lớn nhất của VAR không phải là quyết định sai, mà là những lần hệ thống không quan sát được nhưng vẫn ghi nhận "không có bất thường". **Dữ kiện chính** (3-5 gạch đầu dòng, mỗi dòng ≤25 từ): - 12% trong 240 tình huống việt vị mùa giải 2017 của Shenzhen FC có lỗi căn chỉnh camera. - Kho dữ liệu 1.400 quyết định VAR giai đoạn 2017-2019 là cơ sở cho nghiên cứu công bố tháng 3 năm 2021. - Trọng tài thay đổi quyết định ít hơn 23% khi sân vận động có hơn 40.000 khán giả. - Pha phạt đền của Antoine Griezmann ngày 16 tháng 6 năm 2018 tại Kazan Arena đúng theo góc máy thứ bảy. - Pha phạt đền của Raheem Sterling ngày 7 tháng 7 năm 2021 tại Wembley vi phạm nguyên tắc tiếp xúc tối thiểu. **Nguồn và ngày công bố**: Báo cáo kỹ thuật VAR của Han Chengyu gửi ban tổ chức giải mùa 2017; nghiên cứu về áp lực khán giả công bố trên tạp chí phân tích bóng đá châu Á tháng 3 năm 2021; bản phân tích 5.000 từ về
Minute 73 at the Shenzhen stadium, 2026 season. Shenzhen FC hosting Guangzhou Evergrande. An offside situation looked obvious from the stands, yet the VAR room emitted no signal at all. No whistle. No check. No announcement. The big screen kept showing the score, the match rolled on, and tens of thousands of people believed the machine was running smoothly.
I was sitting in the club's VAR operations monitoring room, and I read a different line: the system reported "no anomaly". That answer kept me awake for many nights. A system can fail in two ways. It fails when it makes the wrong decision. And it fails when it makes no decision at all, then writes into its own log that everything was fine.
After the match I reviewed all 240 offside situations of the season. Twelve per cent of them carried camera calibration errors. I wrote a 30-page report for the league organisers and published not a single line in the media. The following season the positioning system was upgraded. Nobody in the stands knew why. The flaw is not in the system, but in the belief that the system is right.
A machine that never declares its own faults
A VAR room runs on three layers of people. The lead VAR referee holds the final judgement. The assistant VAR watches plays outside the main camera zone. The replay operator rebuilds the situation, selects the frame, draws the offside line. Three people, three screens, one threshold: a clear and obvious error.
That threshold is where everything starts to slip. "Clear and obvious" is not a measurable quantity. It is a judgement, and judgement depends on what a person sees, from which angle, within how many seconds. A referee who sees contact at the ankle from behind can reach a very different conclusion from one who only sees two shoulders collide from the stands.
What interests me more is the data structure the system produces. After every round, organisers publish the number of VAR interventions, the number of overturned decisions, the average review time. Every one of those indicators is positive. They measure what the system did, and no column measures what the system missed.
An empty report can look exactly like a valid one. If the software does not register a play, the log displays "no event". If a camera is off-axis, the offside line is still drawn to the centimetre. No error message appears, because the system was never designed to doubt itself.
240 situations, 12 per cent error, and a gap nobody measures
My review of those 240 situations was fairly manual. For each play I pulled three frames: the frame of the ball leaving the foot, the frame of the drawn line, and the wide frame of the pitch. I compared the vanishing point of the line against the real pitch markings and measured the angle between the camera axis and the vertical axis of the pitch. An error of one to two degrees at thirty metres produces a deviation of fifteen to thirty centimetres on the line. That deviation is larger than the decision margin in most offside situations.
Twelve per cent sounds small. Scaled to a season, it means nearly thirty situations judged by a line drawn in the wrong place. How many of those produced goals, how many disallowed legitimate ones, nobody counted. Because counting them would require admitting that the system might not be seeing at all.
I did not send that report to the press. I sent it to the organisers, with a list of every specific play, its minute, and frame captures. There was no declaration about who was right or wrong. Only data, and one technical question: if the camera axis is off, why did the pre-match calibration procedure not catch it. Four months later I received a two-line reply confirming the system would be upgraded before the 2026 season.
The seventh camera angle in Kazan
On 16 June 2026, at Kazan Arena, France met Australia in their opening group match of the World Cup. The referee initially declined to award a penalty for the contact between Josh Risdon and Antoine Griezmann. VAR intervened, the referee changed his decision, Griezmann converted. France won 2-1.
In the newsroom that day, almost everyone insisted the referee was wrong. The broadcast footage, replayed from the stands angle, showed Risdon touching the ball first and only then touching Griezmann's leg. I asked the technician to pull the seventh camera angle, positioned behind the goal, and that was the only angle showing Risdon's standing leg blocking Griezmann's run before the ball was pushed away. From that angle, the penalty was the correct decision.
The seventh camera angle shows that truth is a relative concept. I spent the next two weeks building a new analytical framework, judging decisions on the basis of what the referee actually saw in real time rather than on slow-motion footage replayed ten times on television. That framework separates two kinds of error that differ in nature.
A technical error occurs when the referee sees correctly but the supporting tool hands him the wrong image. A cognitive error occurs when the image is right but he interprets it wrongly under time pressure. These two errors require entirely different remedies, and merging them into a single word, "wrong", is the fastest way never to fix anything.
Sterling, the minimum-contact principle, and three days of waiting
On 7 July 2026, at Wembley, England met Denmark in the Euro semi-final. In the 104th minute Raheem Sterling went down inside the box, the referee awarded a penalty, Harry Kane's kick was saved but he scored the rebound. An entire nation argued for weeks.
I was the first in my analysis group to identify that this play breached the "minimum contact" principle under the new interpretation of the law. My editor pushed me to publish that same night to capture traffic. I refused, and spent three days completing a 5,000-word analysis of six inconsistent VAR decisions across the tournament.
That analysis became the platform's most-read piece of the year, and it was read most in its third month after publication, not in its first week. That taught me a lesson about the economics of slowness: noise generates traffic for twenty-four hours, while structural argument generates traffic for years.
40,000 spectators and 23 per cent fewer overturned decisions
In 2026 global football stopped. I lost all my broadcast contracts and spent six months building a personal database of 1,400 VAR decisions from 2026 to 2026. I recorded every play: the minute, the score, the attendance, the review duration, the final decision, and whether the referee went to the pitch-side monitor.
A correlation appeared that I had never seen published: referees changed their decisions 23 per cent less often when the stadium held more than 40,000 spectators. In other words, crowd noise does not only affect the referee on the pitch. It affects the person sitting in a closed room hundreds of metres away, watching through a screen.
A database of 1,400 decisions did not find justice, but it found patterns. Justice is a normative concept that cannot be measured in a spreadsheet. Patterns can be measured, and patterns show that people behave differently before the same body of evidence depending on whether forty thousand people are roaring behind them. The research was published in an Asian football analysis journal in March 2026.
Since then, every analysis I write carries psychological context: the timing of the decision, the recent run of matches, the crowd pressure, and even which competition the match belongs to. A decision does not exist in a vacuum. It exists in a room with noise coming through the walls, and in a referee with a heartbeat.
Semi-automated offside and the illusion of precision
Semi-automated offside technology arrived as a promise to end the argument. Dozens of high-frequency cameras, tracking dozens of points on the players' bodies, building three-dimensional models in real time. The line no longer depends on the human eye. It sounds as though the problem has been solved at the root.
But one variable remains that few want to discuss: the moment the ball leaves the foot. The system determines the kick frame through an algorithm, and algorithms carry error. When a striker sprints at nine metres per second, the gap between two frames at fifty frames per second corresponds to nearly twenty centimetres. Choose the wrong frame and the line shifts by twenty centimetres.
The paradox is that the more visually precise the technology becomes, the more easily it creates absolute faith. A line drawn by a three-dimensional model looks far more trustworthy than a line drawn by hand on a flat screen. But faith is not evidence. If the algorithm picks the wrong frame, the line is still sharp, still smooth, still wrapped in beautiful graphics, and still wrong.
What leagues do not publish
Leagues publish the number of VAR interventions. They do not publish camera calibration logs. They do not publish how many plays the system failed to observe for technical reasons. They do not publish the list of situations classified as "checked but no intervention", which is the largest and most opaque group in the entire dataset.
That opacity has practical consequences. A coach cannot know whether a play in the 88th minute was checked at all. A club cannot appeal a situation it does not know exists. And a football culture cannot improve a process that refuses to record its own failures.
Running in parallel is the transfer market, where the same disease appears in a different form. A fee of 222 million euros for Neymar on 3 August 2026 is a verified fact, backed by a contract and a release clause. The thousands of rumours around it have no such backing. Fans read rumours a hundred times more often than facts, and that structure is exactly what keeps trust gaps alive, in the VAR room as much as on the transfer floor.
Crowd emotion and the seventh camera angle
The biggest argument in any VAR controversy was never about the law. It was about the law being right while the feeling was wrong. A technically correct penalty can still be experienced as injustice, because football is not consumed as a legal contract. It is consumed as a story.
This is the counter-intuitive point that took me years to accept. The remedy for VAR is not to make it more accurate. It is already accurate enough that most controversies are arguments about values rather than about data. The remedy is to make it more transparent, to the point where spectators can see exactly what the referee saw, from that same angle, at that same speed.
Modern football is a war between the emotion of the stands and the seventh camera angle. That war does not end with one side winning. It ends only when both sides look at the same frame and both admit that the frame has limits. A good referee is not one who never errs, but one who knows where he erred.
The system, by contrast, has no such reflex. It does not know where it erred, because it was not programmed to know. That is why I hold to the rule of not speaking when data is missing, and why I refuse to publish within twenty-four hours of a major match. I lose a fair amount of traffic, but every analysis I write stands up years later.
What has to change
If I could amend one thing in the system, I would start with the audit log. Every league should publish not only the number of VAR interventions but also the number of plays the system recorded as "checked, no intervention", together with review duration and the camera angles used. This dataset turns invisible gaps into arguable gaps.
The next step is publishing error margins. There is no need to disclose every technical parameter, but the error margin of every offside line drawn should be disclosed. A decision with a fifteen-centimetre error margin should be presented as a decision with a fifteen-centimetre error margin, rather than packaged in three-dimensional graphics as an absolute truth.
The rest belongs to those of us in the media. We need to stop treating every refereeing controversy as an entertainment event to be mined for a few hours. A VAR decision analysed within twenty-four hours produces noise. The same decision, analysed over three days with every camera angle available, produces understanding.
Based on my experience of watching matches across many seasons, I believe supporters are not afraid of complexity. They are afraid of ambiguity presented as clarity. When a system dares to say it did not see, it does not lose authority. It earns trust back.
When systems learn to declare that they did not see, football will enter a different era. Not an era of perfect decisions, but an era of verifiable ones. And perhaps then supporters will argue a little less, not because their passion has faded, but because for the first time they can see exactly what the decision-maker saw, from exactly the angle that person saw it.
