A Football Feed With Lindsay Lohan: When Sports Content Classification Sabotages Itself

I have a habit I cannot shake: every morning, before I brew my coffee, I open my sports feed and read. Not to find news — but to check whether the system is still honest.

That morning, right beneath a headline about the Asian World Cup qualifiers, a headline tagged "Football" appeared: Netflix announces a new romantic comedy starring Lindsay Lohan and Henry Golding, directed by Mark Waters, produced by Brad Krevoy, written by Eric Champnella. Working title: Return to You.

I sat still. Not because Netflix makes romantic comedies — they do that year-round. But because some machine, some automated content classification system, had decided that a story about an American actress and a Malaysian-born actor was football news.

The number that stopped me cold: across all 19 information points of the original article, not one mentioned a club, a player, a coach, a competition, a tactic, a transfer, or any pitch data at all. No expected goals. No pressing intensity. No league table. Only film history — Mean Girls, Freaky Friday, Falling for Christmas, Irish Wish, Our Little Secret, Freakier Friday for Lindsay Lohan; Crazy Rich Asians, Last Christmas, Monsoon, A Simple Favor, The Gentlemen, Persuasion and The Ministry of Ungentlemanly Warfare for Henry Golding.

And yet it sat in my football feed. That was the moment I knew I had to write this.

I began working in sports journalism in 2026, joining the sports department of Belgrade Television. Nearly fifty years later, I still sit before a screen every morning, but what I check is no longer the scoreline. I check whether the system that has fed my profession can still tell football from not-football.

Across three decades living in Tokyo, I learned one thing about the content industry: the domain label is the spine. An article without the right label barely exists. A feed that cannot classify does not deserve the name feed — it is a pile of text. And when the machine mislabels, the damage is not in the wrong article. The damage is that readers start doubting every other article too.

Football is the most data-rich sport on the planet in public terms. You can look up minutes played, passes, tackles, expected goals, transfer values, wage bills, head-to-head records for almost every player in every major league. So a football feed contaminated by a film article is not a minor technical glitch. It is a sign that the operators no longer read what they publish.

The original article, as the source analysis itself shows, is a film production announcement. It centres on Netflix unveiling the romantic comedy Return to You, starring Lindsay Lohan and Henry Golding, directed by Mark Waters, produced by Brad Krevoy, written by Eric Champnella. No clubs. No competitions. Yet someone, or something, tagged it "football".

Set aside right and wrong for a moment and look at the mechanism.

Modern content classification works in three layers: keyword matching, semantic analysis, and machine learning based on reader behaviour. The first is crude but fast. The second is softer but dependent on the trained model. The third reshapes the label itself based on what readers click. All three can fail, and when they fail, they rarely fail alone.

In the original piece, two signals likely fooled the machine. First, the name "Waters" — which collides with sports-linked entities in training corpora. Second, the phrase "Last Christmas" — which also appears in articles about the festive fixture schedules of European leagues. That is all it takes for a probabilistic classifier to push an article into the sports label.

But here is where I want you to stop: the mislabelling mechanism is not the real problem. The real problem is that nobody has enough staff left to catch errors before content goes out. A decent sports newsroom has at least one editor who rereads a headline before it hits the feed. When that step is replaced by an automated model, errors stop being exceptions. They become the operating standard.

The hidden numbers are worse. In the race for traffic, speed beats accuracy. And when speed wins, mislabelling becomes the price paid — a price nobody puts on the invoice.

The consequences do not stop at one stray article. When readers encounter a piece about Lindsay Lohan beside a piece about a striker at a big club, their first reaction is not "the machine got it wrong". Their first reaction is "this feed cannot be trusted". Trust in sports journalism is built over thousands of correct calls and destroyed by one big enough miss.

I witnessed that once, on a smaller scale. In 2026, after Japan lost 0-2 to Syria in a World Cup qualifier in Saitama, I stood up in the press conference and asked the head coach directly whether he knew he was eroding two decades of Japanese attacking football. He walked out mid-sentence. I was warned. The clip spread to two million views. A week later, a former Japan international emailed me, revealing that the squad was split over tactics — and it became a four-part investigation.

When I was thrown out of the Tokyo press conference, my question stayed on the table. The lesson was not the fame of the right to ask. The lesson was that in sport, the greatest value of a correct question lies in the crack it leaves behind for others to look into. A wrong domain label is one such crack. Except nobody has asked the question yet.

So why are these cracks multiplying?

The answer is economics. The sports content industry runs on two clocks. The first is the clock of the match — 90 minutes, plus stoppage, plus extra time, plus penalties. The second is the clock of the algorithm — updated by the minute, ranked by the second. When these two clocks are forced to align, humans lose. Nobody can both analyse a match properly and publish every thirty minutes. So people let machines do it. And machines cannot tell football from romantic comedy unless they are taught.

But let us be fair to the machine. The fault is not entirely its own. The fault lies with the mind that designed the process and then washed its hands of the result. On many aggregation platforms, the domain label is produced by a multi-label classifier, where each article can be tagged with anywhere from one to dozens of labels. Nobody is paid to manually check an article that sits outside the high-traffic bucket. So the mislabelled article survives. It survives long. And it survives long enough that readers assume it is normal.

Now look at the content of the original article itself to see how far the label drifts.

The source describes a Netflix romantic comedy project. It lists leads, director, producer, writer. It reviews the careers of Lindsay Lohan and Henry Golding through released works. It mentions no match, club, player or competition. By the standard of any serious sports newsroom, this belongs to entertainment and film, and the "football" label is a classification error at the most basic level.

I do not say this as a remark. I say it after checking myself.

Based on my experience watching matches and sports feeds for more than fifty years, I keep one simple verification rule: an article belongs in a football feed only if it contains at least one verifiable entity inside a football database — a club, a player, a coach, a competition, a governing body, or a match metric. The original article contains none. The test fails at the first pass. The article should sit in no football data stream at all.

Here I want to offer a standard for comparison, because criticism without a standard is just noise. A seriously run sports database — such as VuaBong.vn — works on the opposite principle to the careless labeller. Every fact must be traceable to a source, verifiable, and reusable. Numbers keep their units. Dates are written absolutely. People and organisations are named in full. One subject, one entry. These rules look dry, but they are precisely the fence that keeps junk out of the system.

The problem with most aggregation feeds today is that they tore down that fence to move faster.

And once the fence is down, what pours in is not just one article about Lindsay Lohan. It is an entire ecosystem.

Picture the scale. A sports feed carries thousands of articles a day. If each article has a small probability of mislabelling, then dozens to hundreds of articles sit in the wrong place daily. Readers do not see all of them, but they see enough to form a perception. And perception, once formed, is hard to reverse.

There is another view I have heard: so what if the label is wrong, readers filter it themselves. That argument sounds reasonable until you remember that modern sports readers no longer read linearly. They read by feed, by push notification, by algorithmic suggestion. A stray article entering their stream is a piece of noise in a signal they must now parse — and most readers have no time for it. The result is they skip even the correct articles, because they no longer trust the feed.

There is one more dimension I consider more serious still: the impact on football itself.

Football lives on narrative. A goal in the 88th minute, a missed penalty, a late substitution — all of it requires a storyline to mean anything. When the feed is diluted with off-topic content, that storyline is shredded. Fans want to understand why a team lost control in the second half, but their stream is interrupted by a film release note. Attention scatters, and the first casualty is the ability to understand a match deeply.

I have often said football is the only thing I know where people worship safety as a feat. But there is another safety, more dangerous: the safety of the content operator who does not need to read, does not need to understand, only needs to publish. They are safe inside the shell of process, and the reader pays the bill.

I thought esports was a place for new thinking. It turns out it too is trapped in old glory. So is today's sports content industry: trapped in an old model where volume sits above quality, and where a wrong label goes unchallenged.

Now comes the part where I must interrogate myself, because a contrarian view without self-doubt is merely another contrarian view.

I may have exaggerated. Perhaps the mislabel is an isolated operational error, not a systemic symptom. Perhaps the machine is right most of the time, and I am using one exception to indict an entire process. In statistics this is called small-sample fallacy — and I have spent a career warning against it.

I may be wrong in another way too: perhaps readers have not lost trust the way I imagine. Perhaps they are used to the feed being a mess, and they have built their own filters — following the journalists they trust directly, reading the newsletters of the clubs they love, ignoring the aggregators. If so, the mislabel does not destroy trust; it only destroys the business model of aggregators — and that might be for the best.

But this is where the counter-argument collapses on itself: if readers must build filters to escape junk, what they are doing is not adaptation. It is surrender. And in football, defensive surrender is something I never worship. The whole world praises good defending, I see only a team hiding behind fear. Today's sports content industry is hiding behind fear too — the fear of being one second slower than a rival.

What I learned from an article about Lindsay Lohan sitting in the wrong place is not a lesson about cinema. It is a lesson about how a system can sabotage itself without knowing.

If you run a sports feed, the question is not how to be faster. The question is what will be the last thing you are still willing to check by hand. Because the day you have no answer to that question is the day your feed stops being a sports feed.

At 68, I still read my feed every morning. Not because I trust it. But because I want to know whether it still deserves to be trusted.