A 'Football' Label on an Ozone Report: When Data Deceives Itself

One weekend morning, I opened a file on my screen in the small apartment in Milan. The first line of the metadata read: "Domain: football." The first line of the content read: "CAMe, ozone, ppb, ZMVM, Fase 1." Two languages. Two worlds. A gap that no analytical model I have built across forty-four working years could bridge.

I sat down, poured a coffee, and read the document from beginning to end. No club. No player. No match. No football metric of any kind. Only a public-service environmental bulletin about an ozone contingency in the Mexico City Metropolitan Area, issued by the Comisión Ambiental de la Megalópolis, accompanied by extraordinary vehicle-circulation restrictions for a specific Sunday.

But I kept reading. Because there is a lesson here — not about football, but about how we, as analysts, can be deceived. And if my profession has taught me one thing, it is this: self-deception is the hardest lie to detect.

Before blowing the whistle, I review myself.

In 2026, inside the VAR room at San Siro, I let a decision slip only because I was afraid of being wrong. At minute 56, Higuaín scored to make it 2-0 for Juventus against Milan. The feed showed the Argentine striker was 0.2 metres offside. I hesitated. I did not recommend a review. Milan lost 0-2. Afterwards, the referee supervisor criticised me in front of the whole team, and I went home with a shame no excuse could ever wash clean.

I spent the following month reviewing forty-seven similar situations. I built a checklist of thirty-seven criteria, ordered them by priority, and committed myself to never again issuing a conclusion without grounds, however safe it might feel.

That episode shaped my method to this day. When I analyse a match, I state the situation, list the data, cross-check, and only then conclude. When I write an article, I build the conclusion on a verifiable chain of evidence, not on the feeling of the moment. And when a file carries a "football" label while its content is an ozone bulletin, I treat that not as a typo but as an event that must be dissected to the bone.

Across forty-four years, I have seen that wrong decisions rarely come from a lack of data. They come from data being mislabelled. A player underrated because the data table assigned him the wrong position. A manager criticised because a metric was tagged to the wrong context. A tactic dismissed as outdated because the model assigned it the wrong purpose. And, in the case before me, an article about air pollution assigned to the wrong domain.

To make the story clear, I need to say a few words about how a sports-news data pipeline operates. Most major sports outlets today no longer employ people to read full raw texts before classifying them. They use automated harvesting systems that scan for keywords, assign preliminary labels, and push the processed texts to editors responsible for each section. The problem lies here: the labelling system does not read content the way a human reads. It counts keywords. It looks for syntactic signals. It cannot distinguish between "football is a sport that needs clean air" and "clean air is needed to stage a football match".

That is why an article about ozone can travel through the system under a "football" label. And that is why analysts like me must remain vigilant: every time data passes from top to bottom, each processing layer has an opportunity to lose a portion of the truth.

Now let us go into the details of the text itself — because I believe an honest audit must be carried out on every line of data, even if the final verdict is "there is nothing here".

The principal subject of the document is CAMe — the Comisión Ambiental de la Megalópolis, the inter-state environmental authority of the Mexico City Metropolitan Area. This body is responsible for issuing emergency measures when air quality in the region drops below safe thresholds. In this specific case, it activated Fase 1 — the first and least severe tier of the response scale. The figures cited are 161 ppb and 157 ppb — measurements of ozone concentration in the air, above the alert threshold and forcing the authority to act.

The concrete action stated in the document is a set of vehicle-circulation restrictions for a specific Sunday — 13 September. This kind of restriction in Mexico City operates on the holograma system (a verification sticker on the windscreen), classifying vehicles by emissions and permitting or banning their circulation by day. The text also contains a public-health advisory: residents should limit outdoor physical activity between 13:00 and 19:00, when ozone concentrations peak.

I read every one of these details three times. And I state plainly: there is not a single football element here. No club. No league. No player. No match. No transfer contract. No league table. No tactical metric — no xG, no PPDA, no key passes, no aerial-duel win rate.

So what is actually happening? It is a labelling error at the ingestion layer. And this is a significant event, not because it concerns football, but because it shows how a data pipeline can collapse at its very point of origin.

Let me use the following sections to analyse three data layers and three failure points of this document, and to draw lessons applicable to football analysis — especially Vietnamese football, where the data system is still being shaped and the risk of mislabelling remains very high.

The outermost layer: label and content

The outermost layer of any file is the label. The label is the first thing people and systems see, and therefore the most likely to mislead. Here, the label reads "Domain: football", while the content reads "environmental alert". This is the most dangerous form of error in the entire data chain, because it is not a missing-data error — it is a wrong-data error. Wrong data is always more dangerous than missing data, because missing data makes you search further, while wrong data makes you believe you have enough.

In football, this is equivalent to a player being assigned the wrong position in a statistical database. He might be a defensive midfielder, but if the table records him as a centre-back, every conclusion drawn about him will be wrong from the root. His tackle count will be underrated. His long-pass count will look anomalous. And if you are a manager looking for a centre-back on the basis of that table, you will make a wrong transfer decision.

I saw this happen in Serie A in 2026, when a mid-table side signed a South American player on figures showing an unusually high aerial-duel win rate. It turned out the club's database had assigned him the wrong position — he was in fact a wide midfielder who drifted inside, not a centre-forward as the table recorded. His aerial win rate was very high, but that was because he almost never contested aerial duels — the denominator was too small to mean anything. The club paid a substantial transfer fee for a player who did not fit the position they needed. And it all began with a wrong label.

Every ruling needs a review, including the ruling of data.

The middle layer: provenance and date

The next layer I must examine is the provenance and date of the document. The text states the day: "Sunday, 13 September". But it does not state the year. This is another type of error — the loss of temporal context. It is less dangerous than a mislabel, but it reduces the reference value of the document to almost nil for long-term analytical purposes.

In football, dating is decisive. A match played in September 2026 carries a completely different meaning from one played in September 2026. The pandemic context, the transfer context, the fitness and fixture context — all change how we read a match. When a document omits the year, we cannot place it in the correct sequence of events, and therefore every conclusion drawn from it carries lower probability.

I have witnessed an analysis of Milan's collapse distorted entirely because it drew on data from a season disrupted by COVID without saying so. When people compared that club's seven-match winless run with similar runs in history, they did not account for matches being played in empty stadiums. The defence lost forty-two per cent of its counter-pressing capacity without the crowd noise driving the press. That is a huge contextual variable, detectable only if you know the year and know what was happening in that year.

The deepest layer: transfer value

The deepest layer is the most important, and it is the one most analysts skip. Transfer value is the capacity of a piece of information to move from one domain to another. Here, the transfer value of an ozone document to football is essentially zero, except for one specific case I will discuss later.

Imagine someone trying to extract a football conclusion from this document. They would say: "High ozone concentrations in Mexico City could affect football matches there." That sounds logically plausible, but it is not supported by any data in the document. The document does not mention any football match. There is no fixture list. There is no club. And most importantly, there is no evidence that football matches were affected. To reach such a conclusion we need an additional data layer — the fixture list, information on whether matches were postponed, medical reports on effects on players. Without that layer, every conclusion is speculation.

And this is where I differ from many analysts. I believe a good analyst is not someone who can draw conclusions from any data. It is someone who knows when to say: "I do not have enough information to conclude." This intellectual honesty is the foundation of any trustworthy analysis.

Football is a game of errors, but the winner is the one who knows which errors are worth making.

Application to Vietnamese football

I have spent many years following the V.League and Vietnamese football at national-team level. And I believe this is an important moment to talk about the data culture here.

Vietnamese football is in a phase of remarkable growth in data infrastructure. Clubs are beginning to invest in statistical analysis systems. Leagues are beginning to collect detailed data on each match. Media channels are beginning to use figures to illustrate their analysis. These are encouraging steps.

But it is precisely in such rapid growth that the risk of mislabelling is highest. Because when new data enters the system, no one has enough experience to audit its quality meticulously. When clubs hire young analysts who are well trained but have not yet spent enough time to understand the league's specifics, the gap between theory and reality can widen. And when media channels race for figures, the pressure to produce conclusions can lower the quality of verification.

I saw this happen in Serie A in the early 2010s, when Italian clubs began importing data models from England and Germany. Those models, however advanced, were built on a sample of entirely different matches. The tempo of Serie A at that time was slower than the Premier League, average passes were lower, and clubs played within a different tactical structure. When those models were applied without adjustment, they produced wrong transfer decisions and unsuitable tactics. It took about five years for Italian clubs to understand that data must be localised before use.

Vietnamese football has a greater opportunity here. Because it is building from the ground up, it can establish a data culture from the start with high verification standards. It can avoid the mistakes the European leagues made — importing models without checking context, trusting labels without checking content, and drawing conclusions without sufficient evidence.

Forty-four years of observing this industry have taught me one thing: the football industry consumes data greedily, but it rarely digests data carefully. When a wrong label enters the system, it does not stop there. It propagates. It influences transfer decisions. It shapes tactics. It generates public opinion. And when analysts like me try to correct it, we fight not only the truth but our own credibility.

No football club is mentioned in this document. That is not a small matter. It is a signal that the pipeline's classification system failed at the most basic layer. And as football data pipelines grow more complex, we need stronger verification systems, not faster classification systems.

The contrarian angle

There is a trend in sports-data analysis today that I want to confront directly: the idea that every document can be mined for football value if you are clever enough to see it. This is a misconception, and it is dangerous.

Whenever I hear an analyst say, "If you read carefully, this document can actually be applied to football," a part of me grows wary. Because in many cases, that is not analysis. That is invention. And invention, however valuable in many fields, can become the enemy of accuracy when applied without limits.

Consider it more carefully. The document on ozone in Mexico City describes a specific event: an air-quality crisis in a specific period, requiring a specific response from a specific authority. No causal structure within it can be transferred to football without additional data. You could say, reasonably, that polluted air may affect the fitness of players competing outdoors. But to turn that into a football analysis, you would need to know: which players? in which match? at what time? under what medical conditions? with what fitness metrics? Without those answers, you do not have an analysis. You have a hypothesis. And a hypothesis, however attractive, is not an analysis.

I know the temptation is great. I have been through it many times in my career. When you have an interesting fragment of data before you, there is an inner craving to turn it into a larger story. But I have learned, through my own mistakes, that this craving must be restrained. Because once you begin to see football everywhere, you will see it even where it does not exist — and that does not make you a better analyst. It makes you a storyteller, not an analyst.

There is an important philosophical difference here. The eclectic school of analysis, popular in parts of sports media, holds that value lies in the ability to connect everything to everything. The systemic school, which I pursue, holds that value lies in the ability to determine precisely what belongs where, and what belongs nowhere at all. The difference is not small. It decides whether an analyst becomes a reliable source of information or a source of entertainment.

And for Vietnamese football, where every piece of information shared can affect hundreds of thousands of fans, this difference has practical weight. Vietnamese fans deserve analyses that are honest, grounded and verifiable — not stories woven from unrelated fragments to produce a feeling of understanding.

My point is not that we should not be creative in analysis. It is that we should be creative within the framework of intellectual honesty. That means: stating the limits of the data, admitting when we do not know, and never turning a hypothesis into a conclusion merely because the conclusion is more appealing.

A thought to carry forward

What, then, should we take from this document? Not a lesson about ozone. Not a lesson about Mexico. But a lesson about data management, quality verification, and honesty in analysis.

For me, after forty-four years in and around this industry, the lesson is this: every time we receive a file, we must check not only its content but its label. Every time we draw a conclusion, we must ask: would this conclusion change if I discovered I had misread one line of data? Every time we share an analysis, we must be sure it rests on a verifiable foundation.

For Vietnamese football, the lesson is that the building phase is the most important phase for setting standards. Clubs, leagues and media bodies now constructing their data infrastructure should devote resources to ensuring that infrastructure is built on the right foundation. Because fixing an error at the root is always cheaper than fixing an error that has propagated through ten processing layers.

I do not trust my eyes; I trust the slow-motion replay. And in this case, the slow-motion replay shows me plainly what the naked eye had skipped: this is not a football article. It is an environmental bulletin wearing the wrong label. And recognising that — rather than trying to find a football meaning inside it — is the true work of an analyst.

The question I leave you, my colleagues in Vietnam, is this: if your data system hands you a document labelled "football", will you read it as an analyst, or as a storyteller? Will you stop at the layer of the label, or will you go deeper into the content to check whether the label is correct?

Because in football, and in analysis, the truth does not lie on the surface. It lies in the deepest layer, reachable only by those willing to review themselves.