The Saqi Nama Exhibition Tagged as Football: A Classification Error and a Lesson in Transfer-Window Noise

A Hamburg Night and a Misprinted Label

The clock on the wall read 2:14 a.m. Outside the window, Hamburg was as quiet as a stadium with no crowd. I sat at the screen doing the work I always do on nights like this: peeling back the layers of a source to see which layer still holds once the night settles in.

The seventh tab in my browser had been tagged "football" by my system. I clicked it. The page opened, and the first thing that hit me was poetry by Allama Muhammad Iqbal. Then the name of a national heritage institute, the names of three artists, and the name of a minister.

Not a single club. Not a single player. Not a single expected-goals figure.

There are numbers that only tell the truth at midnight. That night, what told me the truth was a zero, and it spoke loudly enough to keep me at my desk for another forty minutes.

How I Grade a Source

My job is to assign trust to information before that information can do damage. I do not predict which team wins. I predict whether a line of news deserves to have money placed on it.

I started in 2026, after graduating from the Journalism Academy, writing for Báo Bóng đá and then serving as a correspondent in Madrid for Báo Thể thao Thế giới. Back then I learned the trade standing in the stands, taking notes, and asking myself why the same match becomes two different stories in two different mouths. The answer lay in the anchor point: each person grabs a different detail and builds the story around it.

Thirty-one years later I do exactly the same thing, except now I choose the anchor point with numbers.

My process has four steps. First, extract entities: does the piece mention a club, a player, a competition, a contract. Second, extract verbs: who is doing what, who is promising what, who is paying. Third, extract tense: is the statement something that happened, something happening, or something that will happen. Fourth, extract the source: does the line come from someone directly involved, from someone retelling, or from a press release.

The fourth step is the one most often skipped, and it is the one that decides everything.

During the transfer window, noise drowns signal through a very specific mechanism. A club says one sentence. Three newspapers turn it into three headlines. Six aggregation accounts turn it into six news lines. By the time it reaches the reader, the original sentence is wrapped in ten layers of cloth, and nobody remembers what colour the first layer was.

These weeks, I read about four hundred news lines. The number carrying a direct, signed, dated source is under thirty. The rest are copies of copies.

That is why I always open a working day with the dullest possible question: which section does this piece belong to.

Saqi Nama: When a Painting Has No xG

I went back to the seventh tab and ran all four steps.

Entities: Lok Virsa, the National Institute of Folk and Traditional Heritage; Off-Grid Studios; three artists, Geytee Ara, Lubna Jehangir and Zara Haider Babry; the Federal Minister for National Heritage and Culture, Aurangzeb Khan Khichi; and Lok Virsa's Executive Director, Dr Muhammad Waqas Saleem. Not one football entity.

Verbs: organise, inaugurate, exhibit, recite, praise, pledge. Not one verb belonging to transfer, renewal or negotiation.

Tense: the exhibition opened; the works are hanging; taking the works abroad is only a pledge.

Source: almost the entire content is statements by officials and artists, carried by The Express Tribune. No independent countervailing voice.

Four steps, four times returning the same result.

The exhibition's theme is Saqi Nama, a work by Allama Muhammad Iqbal. The three artists spoke about the difficulty of translating Iqbal's ideas into visual language. Some verses from Saqi Nama were recited at the event and warmly received by the audience. The minister praised the works as being of a high standard, comparable to the finest international pieces, and pledged to show them abroad. Lok Virsa's executive director stressed the goal of connecting the younger generation with heritage.

Stand far enough back, and every heatmap becomes a painting. Tonight I stood far enough back, and I saw exactly one painting: a state-sponsored event, staged by a national heritage institute together with an independent studio, inaugurated by a minister, carrying a message about extending heritage.

And yet my data pipeline filed it under football.

There are two possibilities. One, a failure in the source-classification layer, when the system hits a proper noun that collides with a keyword, or when a source's section tag is set wrongly. Two, a failure in the ingestion layer, when one item sits next to a sports item in the same feed and gets dragged along.

Neither is frightening. What is frightening is what happens next. Had I not sat there for another forty minutes, this piece would have entered my dataset with a wrong label, and stayed there, waiting to be used to teach a model that Iqbal's poetry relates to expected goals.

One mislabel in the ingestion layer can multiply. A single bad record enters a topic cluster, drags ten records in with it, and the whole cluster is then used as training material. Three months later, nobody can trace why the model started suggesting strange things.

I have seen a similar kind of error at a larger scale, and it was not inside a machine. In 2026, Hamburger SV, the club of the city I live in, travelled to Wolfsburg on the final matchday of the Bundesliga, needing a win to stay up. They had 31 percent possession and created 1.35 expected goals against the hosts' 2.10, yet won 2-1 with two goals in the final seven minutes. I went back through all 46 of their matches that season and found they had outrun expected goals by plus 4.2. That figure distorted every pricing model the bookmakers were using. I staked one thousand euros on the survival scenario and published a warning about the market's systemic error.

What I did not write in that piece was that I had almost missed the signal, because in my head HSV already carried the label "a team that struggles". That label was true to the table, and false to the data.

In 2026, when stadiums closed because of the pandemic, my model collapsed quite literally. The variable "crowd pressure", which carried eighteen percent of the weight in my algorithm, disappeared. Ten consecutive bets lost. The Bundesliga draw rate jumped from 24 to 31 percent, and total goals per match fell by an average of 0.4. An empty stadium is a variable no model anticipates.

Three months later I rewatched 120 matches in front of virtual crowds and wrote a rare confession. Since then, every analysis I publish carries a line about environment: home or neutral ground, full stands or empty.

A wrong label is an environmental variable too. Nobody anticipates it, until it ruins an entire dataset.

The Trap of the Promise

Reading the quotes closely, I found a structure familiar enough to make my skin itch.

The minister pledged to take the works abroad for exhibition. The pledge was repeated several times. No budget, no deadline, no venue, no executing body. Nothing to verify except goodwill.

During the transfer window, I read this sentence about twenty times a week, with only the subject changed. The board pledges to keep the star. The agent says his client wants to stay. The club insists the project is progressing well.

These are sentences shaped like data but without the body of data.

I call them empty signals. They take up space in the feed, they take up time in the reader's head, and they carry not one gram of verifiable information. They are not wrong. They are simply not enough.

The promise structure in that culture piece is identical to the promise structure in a transfer item: belief is placed in a pledge, not in a plan.

Probability is not for believing. It is for sleeping with. If you want me to believe a commitment, bring me the contract, the date, the number, and the name of whoever is accountable if it fails. If all you have is words, I file it under noise and go to bed.

A Fault in the Pipeline, and a Fault in the Reader's Head

The mislabelled record is a technical fault, and in principle very easy to fix. I flagged the record for quarantine from the football dataset, and added a rule: if a piece is tagged football but extraction returns no club, no player and no competition, the system must reset the label itself.

But the fault that keeps me awake is not inside the machine.

We human beings also mislabel each other every day, and there is no validation layer to raise an alert.

A young player is labelled a star after one good match, and every match afterwards is read through that label. A thirty-one-year-old is labelled finished after two injuries, and nobody looks at his running data again. A reliable outlet is labelled "just a rumour", while an account that has never once been right is labelled "confirmed".

In my match-rewatch sessions, the thing I have to train most is not reading heatmaps, but stripping labels out of my head before pressing play.

World Cup 2026 taught me that data can be enjoyed like a beautiful match. It taught me something less poetic too: a beautiful memory is a label that sticks very hard. I almost overlooked Croatia because in my head they already carried the label "a team from the smaller tournaments". The PPDA of the Modrić, Rakitić, Brozović trio was 8.7 at the time, the most aggressive pressing figure among the top sides. Data spoke before memory did.

By World Cup 2026, I met that lesson again in Morocco. Achraf Hakimi averaged 11.4 kilometres per match, the most among full-backs, and the team held a PPDA of 9.3, a rare pressing discipline. I backed Morocco to beat Portugal in the quarter-final at odds of 3.2 and wrote a long piece on the data of astonishment. Morocco won 1-0. Had I kept the old label in my head, I would have missed one of the best stories of the tournament.

Data is a temple, and I am only the one sweeping the leaves. My daily work is to sweep away the labels dropped on the floor, not to worship what is already known.

Signals for the Next Round

There is one thing I take out of this Hamburg night, and it belongs neither to football nor to painting.

It is a small rule: before believing a news line, ask which section it belongs to. If the answer is a label, read it again from the top. If the answer is a club, a number, a date, a signature, write it in the book.

The rest of the transfer window will pile into the final weeks, when the noise peaks. Direct sources will thin out, headlines will grow louder, and the daily volume will double. That is when the most valuable skill is not reading fast, but knowing where to stop.

I will quarantine that bad record, fix the classifier, and pour one more coffee. Thirty-one years in this trade have left me with exactly one lesson, repeated over and over: mislabelling something is not what frightens me. Forgetting that I applied the label is.