Dirty Data and Unknown Players: How Mislabeled Feeds Are Eroding Youth Football Scouting

On a Wednesday afternoon, Row B of a provincial stadium held exactly eleven people. I sat third from the left, a battered notebook in my left hand and a stopwatch under my right thumb, counting for a left winger who stands barely 1.7 metres tall and wears number 27. In the first fifteen minutes he touched the ball four times, lost it three times, and made one square run across the opposing right-back at the exact moment his team was losing shape. Nobody in the stand reacted. No camera pointed that way. The ground was empty of spectators, but history was still recording every pass.

My phone buzzed. A scouting data aggregator pushed a notification: Football | Winger | 19 | Plan B. I opened it. Inside were forty-two lines describing a director reworking the ending of a horror film, test screenings for audiences, and contingency footage the studio kept in case the primary version did not work.

No club. No competition. No player.

Someone had labelled a cinema article as football, and the system swallowed it without anyone checking. I laughed. Then I stopped, because I realised this happens every day, everywhere, and in Vietnam almost nobody talks about it.

The biggest problem in Southeast Asian youth scouting is not a shortage of talent or money. It is dirty data being treated as evidence.

Four Sedimentary Layers of a Scouting System

A modern scouting system does not run on human eyes. It runs on four stacked layers, and every one of them can admit rubbish.

The first layer is ingestion. Crawlers sweep thousands of news sites, bulletins, forums, video descriptions and social posts. The second layer is automated tagging: algorithms read keywords, measure frequency, and assign a domain. The third layer is human verification. The fourth is where data becomes a decision — shortlists, valuations, probability models, academy curricula.

In Europe the first three layers have dedicated staff, and the third is usually the most expensive. In Vietnam's V.League 1, with its fourteen clubs, the number of full-time data analysts I could count fits on one hand, and I have asked directly at four different clubs across the last two seasons. Most of that work falls to an assistant coach as a second job, usually after training, usually on a personal laptop, usually with nobody checking it a second time.

Excavating from the sedimentary layer, where names have not yet been carved into legend — except that layer is now being mixed with rubble from an entirely different quarry.

Layer One: Where the Data Is Dug Up

Picture a news website. The lead story that day is about a film. But the navigation bar on the right has a "Sport" tab. The footer carries a live-scores widget. The meta tags contain the word "tournament" because the site covers an annual film festival. A crawler cannot read semantics. It reads structure. And the structure says: this is a football page.

I once sat with a friend who does engineering for a regional sports aggregation platform. He said something I have never forgotten: "We don't classify content. We classify URLs." Once a URL is tagged as sport, everything passing through it carries that tag — including a two-thousand-word film review.

In Vietnam the ingestion layer has one extra variable: machine translation. A great deal of international reporting is auto-translated and republished. The English word "transfer" can mean a player move or the transfer of rights to a work. "Release" can mean releasing a player from a contract or releasing a film. "Draft" can mean a manuscript or a selection process. Translation engines do not distinguish. Neither does the tagger sitting behind them.

Layer Two: The Keyword Trap

This is the part I want to spend the longest on, because it is the root of that Wednesday.

A film article was tagged as football because of four words: franchise, outbreak, Plan B, and the name of a virus. In English-language sport, franchise means a professional team in the NFL, NBA or MLB. In cinema, franchise means a film brand. Outbreak in public health is an epidemic; in sport, an injury outbreak is a cluster of injuries inside a squad, a phrase every club medical department uses. Plan B in film production is a contingency shot; in football it is the alternative shape a coach switches to when the starting plan is read. And a fictional virus name happens to chime with the name of a sports brand.

Four keyword collisions. One wrong tag. Forty-two lines of rubbish entering a system somebody might be using to make decisions.

The frightening part is not the film article. The frightening part is this: if a film article can get in, so can an article about American football, basketball, or an athletics meet. Worse still — a report on a young player in a lower division, mislabelled into a different domain, will vanish from the system and nobody will ever know it existed.

I do not hunt famous players. I hunt the moment they were forgotten. But if that moment is mislabelled, I do not even get the chance to forget it.

Layer Three: The Gap Named Human

Of the four layers, this is the thinnest in Southeast Asia, and the most important.

A decent verification process needs three questions. Does this content mention at least one real football entity — a club, a player, a competition, a federation? Does it contain at least one checkable metric — minutes, passes, touches, a transfer fee? And what tier is the source — mainstream press, tabloid, or an unattributed post nobody answers for?

Three questions. About forty seconds per record. For a data centre processing several thousand records a day, forty seconds multiplied out becomes a cost no mid-table Vietnamese club wants to pay.

So they skip layer three. Not out of laziness. Because forty seconds, times the number of records, times the number of days in a season, is a figure that makes a board ask: how many points does this buy us?

The honest answer is: none, immediately. It only stops you signing the wrong player, hiring the wrong specialist, or building an academy curriculum on a tactical trend measured from dirty data.

I have been on the other side of this story. In June 2026 I was nineteen, interning as a content writer for a fan page in Nha Trang. I watched Iran play Spain and spent the whole piece dissecting a winger for losing the ball seven times in the first half — a number I counted myself from a three-minute clip online. The site's specialist group told me plainly: no basis in real match play. They were right.

The following week I sat through all twelve group-stage matches just to find under-23 players the official statistics had skipped. The first to change my mind was Morocco's Achraf Hakimi, nineteen that summer, then described in most reports as a surplus defender in the squad. Six years later he was one of the most highly valued full-backs in Europe.

The lesson is not whether I guessed right. The lesson is that the data I used that day was not wrong in its numbers. It was wrong in its context. And a system with no verification layer will repeat exactly that error, at greater scale and greater speed.

Layer Four: The Damage at the Far End

Dirty data does no harm while it sits still. It harms when it is used.

Follow a mislabelled record downstream. It enters a scout's shortlist. That scout compiles thirty names, one of which was generated by a system error. He sends the list up. The academy director reads it, sees a profile matching no player in the internal database, and has two choices: ignore it, or spend twenty minutes checking. Twenty minutes, times ten similar records a week, is a time budget nobody has.

Or it travels another route. The rubbish record enters a valuation model. That model is trying to estimate a young player's market value by comparing him with similar cases. If the reference set contains noise, the output is distorted. Not much. A few percentage points. But in a deal where the gap between asking price and offer decides whether a signature happens, a few percentage points is everything.

Every transfer is a stratum. The hurried count the money; the archaeologist reads the era.

And there is a third route, darker than both. Live data from youth matches, collected by the very people standing at the ground, is routinely resold to betting companies as raw feeds. In competitions where no spectators come, no television cameras roll and no reporters attend — which is most under-17 and under-19 football in the region — the only person present is the data seller. Young players contribute data without knowing, without consenting, and without receiving anything. Live data supplied to betting companies is the darkest side effect of sport's digitisation, and youth football is where it is least supervised.

The Purity Trap

Here I have to argue against myself, because that is the only way an observer avoids becoming a slogan.

If we build a strict gate for every record, we filter out the rubbish. We also filter out everything that does not fit a box. A record about a player with no official statistics, no broadcast footage, only a verbal description from a local coach, fails the gate at the first checkpoint.

Yet that is precisely the kind of record academy archaeology needs.

An archaeologist does not throw away the strange layer of soil. He reads it. A good gate is not a fine sieve but a tiered system: what has been verified goes on shelf A, what has not goes on shelf B, and both shelves are used — with different confidence levels, clearly written on the label.

There is a second paradox I consider more important than mislabelling. While we race to clean dirty data, the "cleanest" data of all — real-time event data recorded by someone present at the ground — is being sold to parties with no interest whatsoever in a player's development. The cleanest thing is the most unclean, morally. And the dirtiest thing — a handwritten note about a fifteen-year-old in a provincial town — is often the most honest.

I have one more weakness to confess, and it bears directly on this subject. In 2026, with competitions suspended, I set up a discussion group on overlooked young players, with nine members, all amateur observers. We traded notes for five weeks, then I left for a new project on Brazilian football. One member sent me a single line: "Good at starting things, no idea how to sustain them." He was right. And I realised my error was not starting too many projects. It was treating discovery as the destination, when discovery is only the first step of record-keeping.

That is also the error of most youth football data systems today. They are good at discovery. They are not good at record-keeping.

What I Brought Back from Row B

That number 27, the boy I counted four touches for in fifteen minutes, produced in the seventy-third minute an outside-of-the-boot pass with his left foot that cut through three lines and landed at the striker's feet, and his team scored. No stand roared, because the stand held eleven people. No software recorded a metric for that pass, because the match was never encoded.

That night, at home, I reopened the system and searched his name. No results. The only online record carrying his name was a line in a squad registration list for a youth tournament, with no date of birth, no position, nothing else.

At the same moment the system was still holding forty-two lines describing contingency footage from a foreign studio, tagged as football, ready for anyone to download and cite.

From the substitutes' bench to the spotlight is a dark tunnel. I dig from the side nobody expects.

If you operate any part of a football data system — even a personal spreadsheet — what you need is not another piece of software. What you need is three lines of rules: which source tier does this record belong to, how has it been verified, and how far is it allowed to influence a decision. Three lines. No algorithm required.

And for those who do this work as I do: keep going to grounds with no spectators. But write it down, do not merely discover it. Because if an unknown player exists in no database at all, whether he is good will mean nothing to anyone but us. And if a film article can reach your shortlist while an outside-of-the-boot pass in the seventy-third minute cannot, then the problem is not that the boy is unprepared. The problem is that your system was ready to receive the wrong thing.