Dirty Data and the Labelling Gap: When a Concert Cancellation Slips Into a Football Analytics Pipeline

A data record labelled "football" moved through an analytics pipeline carrying 22 information points. Inside were a singer, a model, some children at an airport, a ticket vendor, and a show pushed back to September 2027. No club. No player. No formation, no passing metric, not a single minute of ball in play.

I read that record twice. The first time I assumed I had opened the wrong file. The second time I understood: the system was not wrong. We were. We built a funnel wide enough to swallow anything containing words, then stamped it "football" and never went back to check.

What kept me there longer than the absurdity was familiarity. I once sat and hand-counted seven turnovers by an Iranian winger in the first half of a 2026 World Cup group game, convinced I was doing professional analysis. It was Iran's 0-1 defeat to Spain, and the player was Alireza Jahanbakhsh. A specialist group told me it had no grounding in match reality. They were right. My mistake that day was an honest one, the mistake of a twenty-year-old learning the trade. The mistakes now flowing through data systems are no longer honest.

The sports industry has digitised faster than it can understand what it holds. Every European matchday generates millions of data points: player positions per hundredth of a second, pressing metrics, distance covered, shot probability. Youth academies, scouts, journalists and betting companies all drink from one data river.

The problem sits upstream, where few look. Before data becomes analysis it must be labelled: who, what, which sport, which competition, which team. That step is largely automated, and that is reasonable on cost. But it creates a fatal gap. A labelling algorithm driven by keywords and surface context cannot distinguish an article about a cancelled music show from an article about a postponed football match, because both revolve around "an event postponed for logistical reasons".

In Vietnam the gap runs deeper. We have very little trustworthy youth football data. V.League has reasonable statistics, but youth competitions, provincial academies and lower-tier matches sit largely outside any recording system. Across the seasons I have tracked, the share of Vietnamese youth matches with detailed data remains very low against scouting demand. With no data, analytics models either stay silent or, worse, invent. And a system forced to invent will quickly learn to grab whatever falls within reach, including a cancelled show in Las Vegas.

Based on my experience watching matches, I believe dirty data is more dangerous than missing data. Missing data tells you that you are blind. Dirty data makes you think you can see.

What was overlooked in this affair has nothing to do with a software bug. What was exposed is the entire trust chain of sports analytics, and that chain is anchored to counterfeit pillars.

How a record like that survives is worth dissecting. It passed the labeller. It passed the deduplication filter. It passed quality control. Three layers, all three failed, because all three were built to check form rather than substance. None asked the simplest question: does this record contain a club? Does it contain a player?

Vietnamese youth football is heading down the same track, only a few years behind. Major centres publish intake information, age-group details, figures on height, weight and minutes played. But most of those figures are published by the subjects themselves. No third party verifies them. A scout sitting in Europe, opening the file of a seventeen-year-old Vietnamese player, will see exactly what the centre wants him to see. Nothing more.

In 2026 I wrote a long piece on a seventeen-year-old midfielder at the AS Roma academy named Emanuele Bove. No one had ever mentioned him in the media. I hand-counted every touch from ageing video files, reconstructing the profile of a player the official system had ignored. The piece drew two hundred reads, but a scout from SPAL — then in Serie B — messaged me asking about my data sources. I was delighted. Then I set up a Telegram group called Youth Diggers with nine members, amateur observers trading notes on forgotten young players. Five weeks later I abandoned it to chase a new project on pressing at Brazilian academies. A member told me: good at lighting fires, bad at keeping them.

I retell that because it shares the same nature as today's story. We are good at generating data and poor at maintaining it. A bad record entering a system does not cause noise once. It becomes the ancestor of hundreds of other records, because the system will use it as the reference for the next one. That is how dirty data breeds: not by intruding, but by being believed.

Unearthing from the sediment, where names have not yet been carved into legend. That is how I define my work. I do not hunt stars. I hunt the moment they were forgotten. But an archaeologist also needs to know whether the layer he is digging is real. If the stratigraphy is mixed, every conclusion drawn from it is worthless, however refined the method. Every transfer is a geological layer. The hasty count money; the archaeologist reads an era. And a poor archaeologist reads the wrong layer, then writes the portrait of an unknown player using data that never belonged to him.

That is why I say this plainly: the biggest problem in modern football analytics is not a shortage of data. It is the sale of live data to betting companies, which I regard as the darkest side effect of sports digitisation. When betting money flows through the same pipe as analytical output, the pressure to be fast will always beat the pressure to be right. And a bad record entering that pipe will be replicated at the speed of light, before anyone asks which sport it belongs to.

The instinctive reaction is to blame the algorithm. I think that is the most comfortable way to dodge responsibility.

The algorithm does exactly what people taught it: prioritise speed, prioritise coverage, prioritise record volume. If the leadership of an analytics system sets a target of one million records a day, then a few thousand junk records inside it is not a bug. It is an operating cost quietly accepted, and nobody wants to say that out loud.

More worrying is our verification reflex. In this case every exculpatory piece of evidence came from one side: the subject himself and his family. Timestamped airport photos, in-cabin photos, an account of two flights that never took off. It sounds persuasive, and may well be true. But no airline spoke. No local promoter confirmed. No third party verified. The whole story stands on the legs of its own narrator.

Vietnamese football lives with this kind of source asymmetry every day. Transfer news comes from agents. Injury news comes from clubs. Form news comes from the players themselves. And we, the writers, often republish it verbatim because of the pressure to be a few minutes faster than a rival. I have done it too. I once published a transfer figure without checking its origin and had to correct it two days later. The lesson was cheap; the reader's trust was not.

From the bench to the spotlight is a dark tunnel. I dig from the side nobody expects. But if that tunnel is drawn with data supplied by the people inside it, I am digging in someone else's darkness, not my own.

The story that began with a cancelled show in Las Vegas will be forgotten within weeks. Tickets are honoured, refunds are offered, a new date is set. Everything closes neatly, and nobody mentions it again.

But the gap that let it slip into a football analytics pipeline remains, and will swallow more. The stands are empty, yet history is still recording every pass — and history only records correctly when the recorder knows what he is recording. I will return to this in six months, when the new season's automated labelling systems come online. If the error rate has not fallen, the question will no longer be whether the algorithm is clever, but whether we have the courage to admit we were lazy.