Mislabeled: When an Art Exhibition Slips Into the Youth Football Data Bin

Two months ago, at 11 p.m. in Nha Trang, I was going through a file of 340 items to prepare an analysis of Japan's U19 crop when I caught a strange line. The item was tagged "football," but the headline was about the opening of the Saqi Nama painting exhibition at Lok Virsa Heritage Museum, Islamabad. I expanded the content. No players. No matches. No metrics. Only three artists — Geytee Ara, Lubna Jehangir and Zara Haider Babry — a Federal Minister for National Heritage and Culture, and verses by Allama Muhammad Iqbal recited at the opening. I read it three times to make sure I was not mistaken. Then I wrote in my notebook: "Football has just been confused with painting. Nobody noticed for weeks." That was the moment I understood that after years of digging up nameless figures, I had stumbled onto another kind of sediment — data sediment contaminated with impurities. To see why this is more serious than a typo, you have to look at how the sports industry runs information in the 2020s. Automated content pipelines push through hundreds of thousands of articles a day, using language models to classify by topic, team and competition. The "football" tag is no longer a word a reader types into a search box; it is a data field that analytics platforms, bookmakers, clubs, and even youth academies such as PVF and Hoang Anh Gia Lai Academy use to filter information. A label error travels straight into training data, into prediction models, into the market value of a young player, before anyone has time to check. Late in 2026, I sat in an online workshop with PVF scouts, after my analysis of how Japan used chaotic pressing to break Germany's order at the Qatar World Cup. One of the scouts asked whether I trusted automated data. I said I trusted hand-counted numbers more. That answer was called old-fashioned. Two years later, holding a file of 340 items with an art article in the football bin, I knew I was not old-fashioned. I was just wary early. Youth football and lower-league football are where data is thinnest. U17 and U19 matches in Vietnam's national competitions, Second and Third Division games, or training sessions at provincial academies have no complete official statistics. Observers like me have to count, record and store everything by hand. When the base layer is thin, a labelling error goes undetected because nobody cross-checks. That is why an article about an art exhibition in Islamabad sat quietly in the "football" bin for a whole month without being questioned. Across the item's 34 information points, none relates to football. The exhibition is titled Saqi Nama, inspired by the poetry of Allama Muhammad Iqbal. It was organised by Lok Virsa — the National Institute of Folk and Traditional Heritage — in collaboration with Off-Grid Studios. It was inaugurated by Federal Minister for National Heritage and Culture Aurangzeb Khan Khichi. It was run by Lok Virsa Executive Director Dr Muhammad Waqas Saleem. Three artists took part: Geytee Ara, Lubna Jehangir, Zara Haider Babry. The content consists of visual works inspired by Iqbal's poetry, a recitation of selected verses, and the minister's pledge to exhibit the works internationally. The minister said the works were of a high standard, comparable to the finest international works. The Lok Virsa director said the exhibition aimed to connect the younger generation with heritage. The verses were recited and warmly appreciated by the audience. All 34 information points sit inside this chain. No line, not even a phrase, can be assigned to football. So the real question is not why this article was mislabelled, but how many other articles are mislabelled without my seeing them. Out of 340 items I scanned, only one was about painting. But I cannot verify whether 3, 5 or 20 similar pieces sit deeper in folders I have not opened. Information asymmetry in the sports data industry is not about missing data. It is about wrong data that is not caught in time. I once thought this was a minor technical issue, until I remembered June 2026. I was 19, writing for a football fan page in Nha Trang, and I built a sarcastic analysis of Alireza Jahanbakhsh after Iran lost 0-1 to Spain at the Russia World Cup. I hand-counted seven losses of possession in the first half from a three-minute YouTube clip, then concluded he was useless. A week later, the fan page's editorial team said the piece had no real-match basis. I was embarrassed, but that was the first time I understood that casually counted data from unverified sources is a form of mislabelling — labelling a player "useless" when I had never watched him for a full 90 minutes. Eight years later, I see the same mechanism replaying at industrial scale. An algorithm does not watch 90 minutes. It reads the headline, extracts keywords, then labels. If the word "heritage" sits next to a sports section on the same newspaper page, the filter can slip. I am no longer 19 and blaming a YouTube clip. I am 27, old enough to understand that a system error is many times more serious than a personal one, because it spreads faster and is caught later. In my academy archaeology work, I always tell younger colleagues that every name must be dug up at least twice from two different sources before writing. That is the minimum rule. But that rule cannot be applied to automated datasets, because nobody knows the true origin of a label. The string "Saqi Nama" carries no association for the algorithm with poetry or painting; it is just a set of characters that needs to be placed into an existing bin. And the nearest bin, by some criterion I cannot verify, was "football". What worries me most is not an art article being mislabelled. What worries me is a nameless player being mislabelled in the opposite direction — pushed out of the football bin because the algorithm cannot recognise him. A U15 player in Dak Lak scoring at 4 p.m. on an artificial pitch, if no journalist writes about him in the first 24 hours, will not exist in any dataset. At that point, being mislabelled is no longer an error. It is disappearance. The most counterintuitive thing is that the sports analytics industry has no wish to fix labelling errors. An art article slipping into the "football" bin costs nobody money, nobody a job, no platform's share price. On the contrary, the swelling of auto-labelled datasets — impurities included — is exactly what gets sold to betting companies, data investors and transfer brokers. Mess benefits the seller. It makes a dataset look fuller, more comprehensive, covering every topic. Nobody audits a contaminated data field until a prediction model returns an absurd result. I am not against automation. I am against automation used to replace verification. At Vietnamese youth academies, a local scout can tell you the name of a U15 player who scored in the 89th minute on an artificial pitch at 4 p.m. — information any algorithm skips. But when data is auto-aggregated and resold abroad, that name vanishes, while an art exhibition in Islamabad surfaces in a "young talents to watch" list. Four years ago I spoke on Telegram with a group of European scouts. They use automated data to screen Asian football. They believe the algorithm is more objective than the human eye. What they do not know is that behind every auto-labelled data field sits a layer of distorted sediment that has never been dug up. When that layer grows thick enough, it buries the very nameless players they are trying to find. I am not writing this to indict any particular platform. I am writing to mark a small event that should have been a bell. The stands are empty, but history is still recording every pass — and if the recorder is an algorithm that cannot tell painting from football, then the real passes will be buried under wrong labels. In six months I will return to this story with a concrete list: how many art articles, how many food articles, how many fashion articles sit inside "young talent" datasets. If that number is greater than one, it is no longer an error. It is a method.