The Wrong Label at the Edge of the Frame: A Misclassification Case and What It Teaches About Reading Sports Data

In the small editing room in Chengdu, I keep a habit that has become professional reflex: before concluding anything about a situation, I pull the frame all the way to its edge. Thirty years of watching football through a screen taught me that truth rarely sits at the centre of the frame. It sits at the rim, where few people look. In 2026, when I worked as a VAR data analyst for a sports channel, I reviewed 147 controversial refereeing situations across 28 rounds of the Chinese Super League. Twelve incorrect offside decisions were directly linked to camera placement. The match between Guangzhou Evergrande and Shanghai SIPG in round 25 is the one I remember best: Wu Lei's goal was disallowed for a 15-centimetre deviation, while not a single camera was positioned on the correct horizontal plane to establish the offside line. A 15-centimetre error was confirmed by an angle incapable of measuring it. From there I built my own database of viewing-angle error, logging every incident, every camera position, every margin of deviation. That work did not let me re-referee a finished match. It taught me something larger: the frame you are shown determines the conclusion you are forced to reach, and most viewers forget they are only being shown one frame. I found the error not at the centre of the pitch, but at the edge of the frame. This summer, another frame raised the same problem, except it was not on grass. A news item entered a sports data pipeline, tagged as belonging to football, while inside it there was not a single club, player, match, coach or football-governing body. It was an open criminal case in Atizapán de Zaragoza, State of Mexico, concerning the family of a musician — the keyboardist of the band Camilo Séptimo — along with two suspects whom the source describes as vinculado a proceso. Under Mexican criminal procedure, that phrase means the accused has been bound over for trial, not convicted. I am not writing this to retell a family's tragedy. That belongs to a court, to the people who have suffered a loss, and not to a sports page. I am writing for a narrower but more uncomfortable reason: that misapplied label is an offside at the edge of the frame, and it exposes how we — data people, referees, journalists — read the world. When the data pipeline misreads a name To understand how a crime report can wear football's shirt, one concept needs stating clearly. A data pipeline is the automated sequence a piece of content travels through before it reaches a reader: collection, entity extraction, topic tagging, section routing. Entity extraction is the step where a machine finds names of people, organisations and places in a passage. Topic tagging is the step where it decides whether the item belongs to football, music or the judiciary. Both steps rely on patterns, and patterns are blind to context. A string of characters matching a familiar entity is enough for the machine to apply a label. A band name, a nickname, a place name, a job title — anything matching the dictionary the system uses — can drag an entire article into a strange section. The machine does not read meaning. The machine counts matches. In VAR analysis we call this the false-correlation pattern: two things co-existing in a frame does not mean they are causally related. A defender raising his arm at the moment the ball touches it does not create an offence if the arm is in a naturally defending position, but if you only see those two events side by side, your brain links them into a story. A tagging algorithm behaves the same way, except it has no brain with which to doubt itself. The scale matters. In a modern publishing system, thousands of items pass through the pipeline each day. A mis-tagging rate of even one per cent produces dozens of errors daily. Most are caught at the editing desk. A small number slip through, and the ones that slip through are often the most dangerous, because they touch sensitive subjects — where an error does not merely pollute data but harms real people. It took 37 reviews before I understood that the human eye is not a measuring instrument. I said that about a football incident, but it holds for every automated reading system. The human eye, and the machine eye, are both instruments with error. The question is not whether error exists, but whether we know where in the process it sits. When a crime report is filed under football, that is not a catastrophe. It is a signal. It reports that somewhere in the pipeline, one step is looking at the edge of the frame and believing it is looking at the centre. Nine dimensions, and the only one that mattered When I received the request to analyse that item across the usual nine dimensions — tactics, transfer finance, results and public opinion, league context, rules and governance, dressing room, risk profile, media narrative, industry transmission — I did exactly what I do in a VAR room: I checked each dimension and said plainly which ones had no data. Tactics: none. No formation map, no passing metrics, no pressing model. The central subject of the item is a musician, not a footballer. Camilo Séptimo is a band, not a club. Finance and the transfer market: none. The only commercially coloured element is an investigative thread about a business relationship between the musician and a suspect. That is a judicial matter, not a contract, wage bill or financial-fair-play question. Results and public opinion: none. No table, no form, no pressure on a manager. The only thing resembling public opinion is public interest in a criminal case with a celebrity element — a different kind of attention from sporting attention. League context: none. The only context is geographic and judicial — the events occurred in Atizapán de Zaragoza, State of Mexico, and are handled by a specialised prosecutor's unit. Football rules and governance: none. No FIFA, no federation, no competition regulation. But here a transferable layer appears, and I will address it in detail below. Dressing room and management: none. The nearest analogue is a band losing a member. That is a cultural event, not a sporting one. Transmission into the football industry: none. There is no path from this item into the football ecosystem. Any attempt to draw one would be fabrication. So what remains? Two things, and neither is football. The first is the risk profile, at the level of journalism. The item rests almost entirely on a single source — the family's lawyer. In professional practice, a single source is an unverified source. Not because the lawyer is lying. Because any legal representative has a natural interest in shaping an emerging narrative. That is the nature of representation, not an accusation. The second is media narrative, and this is the only dimension with genuine analytical value. The item is a clean example of negation-framed reporting. The phrase describes a story whose central "news" is the absence of a fact — here, that tests found no intoxicating substances in the victim's body. Such items suit newsrooms well, because they hint at an explanation for the motive question without making any legal conclusion. But the source itself states clearly that the case remains open, that the timeline is not definitively fixed, and that other expert tests are still pending. In other words, the item sets its own marker: it admits it has established nothing. I found the error not at the centre of the pitch, but at the edge of the frame. We think we are searching for justice, when in truth we are only searching for a prettier camera angle. With negation-framed items, the prettier angle is the one that lets readers complete the story themselves without anyone having to assert anything. One source, one negation, one frame Here I need to state clearly why I refuse to produce football analysis for this item. Not because I lack the imagination to build an interpretation. I could easily write five hundred words arguing that a band losing its keyboardist is equivalent to a team losing its midfield metronome. But that is the kind of interpretation that produces a feeling of depth without producing knowledge. It is like a commentator saying the away side "lacked hunger" when the home side had 68 per cent possession. It sounds insightful, but it explains nothing. My craft is tracing decision chains. For a football incident, the chain includes time, position, camera angle, player movement and the perceptual limits of the referee. For a news item, the chain includes the source, timestamps, framing structure and the limits of the reader. In this item, the decision chain is very short. It starts with an interview, passes through a lawyer's hands, reaches a newsroom, is pushed into a data pipeline, and at the end of that pipeline an automated tagger decides it belongs on the football page. Not one step in that chain checked whether the subject of the story had any real connection to football. That is the genuine parallel with VAR. A VAR room receives one camera angle, and the first question should always be: is this angle capable of answering the question being asked? If the offside line falls outside the camera's coverage, the honest answer is that no conclusion can be drawn. Not offside, not onside. But: I do not have sufficient basis. A decent data pipeline must have exactly that reflex. When the central subject of an item does not belong to the field it has been tagged into, the correct reflex is to remove the tag, not to force the content into the frame. Forcing content into a frame does not merely create dirty data. It also delivers an open criminal case to a sports audience that is not prepared to read it correctly. That is two harms at once: a harm to the data, and a harm to the dignity of those involved. In this case, the two suspects are referred to by the convention of reducing their surname to the letter N. That convention exists to protect the presumption of innocence, and anyone republishing the story must preserve it. This is a small detail of judicial technique, but a large detail of journalistic ethics. The presumption of innocence is not a procedural formality. It is a binding standard before any verdict exists. Presumption of innocence and the "clear and obvious" standard Under the laws of football, VAR may intervene in only four categories of situation, and only for a clear and obvious error or a serious missed incident. The phrase "clear and obvious" is a deliberately vague clause. Its vagueness is not a drafting error. It is design, because a threshold set too low turns every incident into a meeting, and a threshold set too high turns VAR into a meaningless ritual. I am sceptical of that threshold. But I respect the principle behind it: intervene in a decision only when the basis is overwhelming, and otherwise leave the original decision standing. At the level of journalism, the equivalent standard is the presumption of innocence. Before a verdict, every assertion about an accused person's conduct must clear the highest intervention threshold. The counter-intuitive angle sits here. The item that draws the most attention is not the one saying the case is open. The item that draws the most attention is the one hinting at an explanation for the accused's conduct by ruling out another explanation. Saying no intoxicating substances were found in the victim's body sounds like a neutral fact. But placed at the centre of a story, it becomes an interpretive frame. The reader fills the gap with a question: if not that, then what? And an answer appears before any court has said anything. This is what I call seeing with the eye and still not being right. In the Perisic situation at the 2026 World Cup final, I measured six main broadcast angles, and only one showed the arm extended in a way judged unnatural. That angle was the only one the referee was shown in the VAR room, for one minute and forty-seven seconds. Six angles, one chosen, one decision produced. No one lied. A frame was simply placed at the centre while the other five were pushed to the edge. The empty stadiums of 2026 showed me this: VAR does not save football, it exposes football. When there is no crowd noise to fill the silences, you hear the referee's exchange with the VAR room clearly, and you see that technology does not create justice. It only creates a record. Whether that record is full or empty depends on which angle someone chooses. The crime report tagged as football is one such record. It has one source, one negation, one frame. What does not appear in the record — the prosecution, the defence, the unfinished expert tests — is pushed outside the frame. No one deleted them. They simply were not shown. And here the "clear and obvious" threshold becomes useful. Apply it to this item and you are forced to ask: is the central assertion clear and obvious? The answer is no. The source itself says everything may still change. So the intervention threshold is not met. An honest writer must leave the original state standing: the case is open, the accused are not convicted, the motive is not established. Reading the edge of the frame Thirty years in this trade taught me that most disputes are not about the event. They are about the boundary of what was recorded. Referees are not wrong when they fail to see what their camera angle hides. Readers are not wrong when they believe what the data pipeline led them to. The error lies in a system that placed one frame at the centre and called it the whole picture. In this transfer window, as rumour noise drowns out signal, the most valuable skill is not predicting correctly. The most valuable skill is telling a fact apart from a frame. A published fee is a fact. A name attached to a club is a frame. A test result is a fact. A motive story built around it is a frame. The fix is not to trust less. It is to demand more angles before concluding. For an offside line, that means a camera on the correct horizontal plane. For a news item, that means a second source, a third source, and a professional reflex cold enough to say: I do not have sufficient basis. The wrong label at the edge of the frame will be corrected. A new tagger, a new rule, a routine audit — the pipeline will run clean again. But the frame stays far longer. Readers have now seen the story in a certain shape, and that shape is harder to erase from memory than a line of metadata. I write these lines in the editing room, still with the old habit: pulling the frame to its edge. Thirty years from now, I will probably still do it. Not to find the error at the centre of the pitch, but to remember that the centre of the pitch is merely where the frame is pointed, not where the truth has been placed.