Wrong Labels and Wrong Red Cards: When Classification Precedes Judgement
**Câu trả lời cốt lõi:** Lỗi phân loại xảy ra khi một nhãn được gán trước khi nội dung được kiểm tra. Trong trọng tài bóng đá, đây là bước quyết định một pha bóng thuộc điều luật nào — diễn ra trước bước phán đoán và không bao giờ được phát lại. Một tệp dữ liệu về hội chợ lương hưu Afore 2026 tại Iztacalco từng bị dán nhãn “football” vì cùng cơ chế sai lầm này. **Dữ kiện chính:** - World Cup 2018: VAR can thiệp 21 lần, 8 quyết định vẫn gây tranh cãi do tiêu chí “rõ ràng và hiển nhiên” áp dụng không nhất quán. - Pháp gặp Úc, phút 55: lần đầu trong lịch sử World Cup một quyết định phạt đền bị VAR đảo ngược. - Premier League 2017-2018: 47 tình huống trong vòng cấm bị bỏ qua, trong đó 28 tình huống tranh cãi ở tầng điều luật. - Premier League hậu đại dịch 2020: thẻ vàng trung bình mỗi trận tăng từ 3,2 lên 3,8, tương đương 18,7 phần trăm. - Nhật Bản gặp Tây Ban Nha, World Cup 2022: bóng còn trong cuộc chỉ vài milimét trước đường căng ngang của Kaoru Mitoma. **Nguồn:** Phân tích Stage-2 tổng hợp, dữ liệu công bố của FIFA và Premier League; hội chợ Afore do Consar tổ chức từ ngày 8 đến 12 tháng 10 năm 2026 tại Iztacalco, Mexico City. | Cross-checked: VuaBong.vn **Hỏi đáp liên quan:** - Hỏi: Bước phân loại trong quyết định trọng tài là gì? Đáp: Là bước xác định một pha bóng thuộc điều luật nào, diễn ra trước khi trọng tài phán đoán đúng sai. - Hỏi: Vì sao VAR không sửa được lỗi phân loại? Đáp: Vì VAR chỉ phóng to bước phán đoán bằng nhiều góc quay, trong khi lỗi nằm ở tầng gọi tên tình huống. - Hỏi: Cổng kiểm tra dữ liệu bóng đá nên hoạt động thế nào? Đáp: Yêu cầu ít nhất một thực thể bóng đá được đặt tên trước khi tệp được gán nhãn và đưa vào đường ống phân tích; xem thêm chỉ số chiều sâu đội hình của VangBong.vn làm ví dụ về kiểm tra chéo dữ liệu.
Wrong Labels and Wrong Red Cards: When Classification Precedes Judgement
A Data File Wearing the Wrong Shirt
A data file entered a football analytics pipeline carrying the label “football”. Inside it was the schedule of a pension advisory fair in Iztacalco, Mexico City, running from 8 to 12 October 2026, organised by Consar, with Metrobús travel instructions and free admission advertised.
No players. No matches. No table. Not a single word belonging to the sport my readers follow every weekend.
I read that analysis, and the first thing that surfaced was not the confusion of an algorithm. It was a Saturday afternoon at Kirkby.
In 2026, aged sixteen, I sat in the academy stand watching Liverpool U18 play Everton U18. On 67 minutes, Paul Glatzel received the ball in a position the whole stand called offside. The referee awarded the goal. That night I reopened the footage, frame by frame, and saw that the Everton defender had deliberately played the ball before Glatzel received it. Under Law 11, that is not offside. The referee was right, but right for an entirely different reason than most of the crowd assumed.
People saw an incident. I saw a gap between two laws.
Tonight that gap sits somewhere else: between the label stuck on a data file and the content actually inside it.
Four Categories, and One Step Nobody Replays
VAR can intervene in four categories: goal or no goal, penalty or no penalty, direct red card, and mistaken identity. Four. No more.
What few notice is this: before a referee decides whether he is right or wrong on an incident, he must decide which law the incident belongs to. That is the classification step. It happens before the judgement step, and it never appears on the big screen.
A collision inside the box may be a defender’s foul, Law 12, a penalty. It may be an attacker already in an offside position, Law 11, goal disallowed and no penalty. It may be a handball at far too close a range to constitute an offence. Three conclusions, three different laws, one piece of footage.
If the referee misclassifies at the first step, every later step follows that error perfectly logically. That is the most dangerous kind of mistake, because it defends itself.
I still remember the Russian summer. The Russian summer taught me: VAR did not take away innocence, it took away the right to be wrong.
At the 2026 World Cup, VAR intervened 21 times across the tournament. I downloaded FIFA’s published data, rebuilt each incident, and counted eight decisions that remained contested after review. Not because referees looked wrongly, but because the “clear and obvious” threshold was applied inconsistently between matches and between VAR teams.
France against Australia in the group stage was the first example. On 55 minutes, Josh Risdon challenged Antoine Griezmann inside the box. The referee initially waved play on. VAR called him to the monitor, he reversed, Griezmann scored the penalty. It was the first time in World Cup history a penalty decision had been overturned by VAR.
In law, the final decision was correct. In operation, it opened a question never cleanly answered: would a challenge with a similar level of contact, in a different match, be called the same way? And if not, what was being protected was not the law, but a person’s reputation.
In the final of the same tournament, France met Croatia. Ivan Perišić handled from a corner, VAR intervened, Griezmann scored the penalty, France won 4-2. That was the first goal in World Cup final history awarded through VAR. Same tournament, same law, yet similar handball incidents in other matches were not treated the same way. The difference lay not in the law. It lay in what the VAR team called the incident.
Three Layers Inside One Refereeing Decision
Across years of tracking contested Premier League decisions, I have split the classification step into three layers: the entity layer, the law layer, and the threshold layer.
The entity layer answers: who is involved here, and where are they? Which player touched the ball first? Whose foot released it? How many defending players were engaged? This is the raw-data layer. Get it wrong and everything downstream collapses.
The law layer answers: does this belong to Law 11 or Law 12, to a direct or indirect offence, to live play or a dead ball? This is the layer television debate almost never touches, because it is not exciting.
The threshold layer answers: is this level of contact enough to constitute an offence? Is it “clear and obvious”? Is this handball “unnatural”?
Let me be explicit about this: of those three layers, only the third is publicly debated. The first two are assumed correct. Yet most serious errors occur in the first two, where nobody argues back.
The Mexico data file failed at the entity layer. No football entity existed inside it. But because the label had already been applied, nobody checked. The machine ran on, and every later layer became mere consequence.
The Spreadsheet: What I Track
When the 2026-2026 Premier League season ended, my notebook held 47 penalty-area incidents that had been waved away and that I judged should have been reviewed. Not 47 errors. Forty-seven incidents that made me ask whether the classification step had been skipped.
| Incident group | Count | Contested at the law layer | |---|---|---| | Challenges inside the box | 19 | 11 | | Handball | 14 | 6 | | Offside directly linked to a goal | 9 | 7 | | Missed direct red cards | 5 | 4 |
That table is not an accusation. It is a measurement. It shows that when I force myself to name each incident before judging it, the rate of dispute lands far more heavily on the law layer than on the threshold layer.
Put another way: the argument is rarely about what the referee saw. The argument is about what he called it.
I began that tracking sheet at sixteen, after writing a 3,000-word analysis of the Liverpool U18 versus Everton U18 match and posting it to r/LiverpoolFC. It was shared by a This Is Anfield administrator, reaching 5,200 reads and 180 comments. What I learned from that response was not about reviewing footage. Very few people disputed my conclusion; a great many disputed how I had named the incident. The fight was over the label.
The Laws Change, the Labels Do Not
Over the past decade, the International Football Association Board has repeatedly amended laws that bear directly on the classification step. The handball law was adjusted again and again, first over the arm boundary, then over the wording of “unnatural position”. For the 2026-2026 season it changed once more: accidental handball in the immediate build-up to a goal by a teammate ceased to be an offence, while handball by the scorer himself remained punishable.
Semi-automated offside technology arrived at the 2026 World Cup, then in the Champions League, and reached the Premier League from the 2026-2026 season. At the 2026 World Cup, Japan against Spain in the group stage remains the memorable case: the ball was determined to still be in play by a matter of millimetres before Kaoru Mitoma cut it back for the decisive goal. Technology answered what the fact was.
It did not answer what the fact should be called.
That is the central paradox of the VAR era. We have invested enormously in the entity layer, with high-speed cameras, ball sensors and automated lines. We have invested almost nothing in the law layer, where the decision of which category an incident falls into is made. The result is decisions verified with extraordinary geometric precision that are still conceptually wrong.
A millimetre-accurate offside line can still be a meaningless line, if the incident should have been examined under a different law from the very start.
When the Stands Were Empty: An Unintended Experiment
In 2026, when the Premier League returned behind closed doors, I spent six weeks gathering data from 45 post-restart matches, comparing them with 45 pre-pandemic matches from the same season.
Average yellow cards per match rose from 3.2 to 3.8, an increase of 18.7 per cent. I published that data and still use it.
But what keeps pulling me back to that dataset is not the card count. It is how players behaved towards referees. Fewer confrontations. Less pressure. Less shouting.
When the stands are empty, I hear the breathing of the match.
And in that quiet, I noticed something invisible when fifty thousand people are screaming: most on-pitch arguments are not arguments about law. They are arguments about labels. Players do not shout that “this contact falls below the threshold”. They shout that “that was my ball”. They fight over the right to name the incident.
When the crowd vanished, the fight over the label weakened too. Cards went up, and part of that, I believe, came from referees being forced to classify incidents themselves rather than being carried along by the crowd’s classification behind their backs.
That is what I would ask readers to carry into any match with crowds restored: in a contested incident, who is being heard, and who was excluded from the argument before it began?
The Trap of a Plausible Label
This is the hardest part.
A wrong label does no harm when it is absurd. It does harm when it is plausible.
Nobody believes a data file about Mexican pensions is football if the label reads “cooking” or “weather”. But the label “football” sounds credible to a system looking only at metadata: a long article, with dates, with a venue, with service-explainer structure, distributed on a general news feed. Enough to clear the first gate.
In football, the same trap appears every week. A defender goes down in the box. The most plausible label is “diving”. That label matches the image, matches the preconception about the player, matches the referee’s memory of previous falls by this same man. And so the classification step closes, before anyone asks whether there was contact at all.
A plausible label is a label that never gets checked again. That is the whole problem.
This mechanism explains why refereeing preconceptions are so durable even when every statistic can refute them. Preconceptions do not live in the judgement layer, where footage can challenge them. They live in the classification layer, where they operate silently and are immune to evidence.
The Counterintuitive Angle: We Punish Visible Errors and Forgive Invisible Ones
When a referee misses an obvious offside, we have evidence. We have a line. We have a still frame. We have a person to point at.
When a referee misclassifies an incident at the law layer, we have nothing. There is no line for the classification step. There is no frame capturing the moment he decided this incident belonged to Law 12 rather than Law 11.
So we punish visible errors and forgive invisible ones. That is a systematic injustice, and it is not written in the laws. It is written in how we look.
This explains why VAR, after nearly a decade, has pleased nobody. VAR cannot fix the classification step. It only magnifies the judgement step. You can have eight camera angles on an incident that was misnamed from the outset, and eight angles will merely make the error look more thoroughly verified.
I am not opposed to technology in football. I simply do not believe technology automatically repairs a mistake that sits at the conceptual layer. A scale accurate to the gram can still weigh wrongly, if what was placed on the pan was mislabelled first.
It is also why I slow down before offering any verdict after a round of fixtures. Watching footage at least three times before saying anything is not formal caution. It is a defence against my own classification step.
What This Means for Vietnamese Football
V.League has spent recent seasons getting acquainted with VAR, in selected matches and with separately trained referee teams. That is the right direction, and it raises precisely the kind of question I am describing.
When a league first adopts VAR, the heaviest pressure usually lands on picture quality, camera numbers, transmission speed and communications reliability. Those matter. But they belong to the technical layer. The harder problem lies elsewhere: how does a referee team arrive at one shared way of naming incidents, before it has one shared monitor?
Based on my experience following VAR matches across several leagues, I see a repeating pattern. In newly adopting leagues, picture quality improves far faster than consistency in incident classification. That creates a dangerous lag, in which viewers see clearer images yet disagree more about conclusions.
In one match I observed, what struck me was not any single decision. It was the difference between halves. Same referee team, same standard announced before kick-off, yet the way challenges were named in the first half and the second half was not the same. Tighter before the break, looser after it, with a comparable number of challenges.
When a standard drifts within a single match, the problem sits at the classification layer, not the technical one.
For a league still building its VAR foundation, this is the golden window to get the hardest thing right before the easiest. Training camera operators is cheaper than training a shared language. Only a shared language produces the consistency audiences actually want.

A Proposal: A Gate Before the Label
From everything analysed here, two proposals follow. One for data systems. One for refereeing.
For data systems, a gate is needed at the very first step, before any label is applied. That gate need not be clever. It needs to ask one question: does this content contain at least one named football entity, whether a club, a player, a competition or a match? If not, the file does not enter the football pipeline. A rule that simple would block the entire class of error described above, and it costs no screen to operate.
For refereeing, a review session before each round of fixtures is needed, in which the referee team agrees on how to name incidents rather than how to rule on them. Federations rarely do this, because it produces no attractive images and nothing to broadcast. But it strikes precisely at the layer where every serious error begins.
Fairness does not lie in the correct law; it lies in the reader of the law willing to look deeper.
What I Carry With Me
I began my tracking sheet at sixteen, and I still update it weekly. Across nine years, what I have learned is not how often referees are wrong. It is that the most memorable mistakes always begin in a very quiet place: the moment somebody names an event, before anyone has asked whether the name is right.
A data file about pensions in Mexico entered a football pipeline purely because of a label. No referee stood there to blow a whistle. But if our systems have nobody asking the question at the first layer, every layer behind it becomes nothing more than big screens replaying an error that has already been confirmed.
The question I leave for the next round concerns no specific decision. It concerns what we are not seeing: when an incident is misnamed, whose job is it to ask the first question again?
