When an Algorithm Tagged an Electricity Audit as 'Football'

I once mispronounced Luka Modrić's name three times in one half of a 2026 World Cup semi-final on Russian soil. Three times, in front of a national broadcaster's microphone, in front of millions of viewers. I said 'Mo-drich', with the dry English 'ch', when the name must be spoken with the soft Croatian 't'. That night, the station's phone lines would not stop ringing. Viewers called in to scold me, and they were right to.

But my mistake was the mistake of a human who could fix it. I rewatched every tape, relearned every syllable, and within a month I was pronouncing the names of 736 players at that tournament correctly. I once got a legend's name wrong, to learn that football does not forgive carelessness.

This week, I ran into a different mistake. This time, the culprit was not a 51-year-old commentator fumbling at a microphone. The culprit was a machine. An automated content-labelling system tagged a straight news report about Pakistan's power sector with two words: Football.

There is not a single player in it. No goal, no tactic, no transfer, no competition named. Only distribution companies, a technical audit, and line-loss figures. Yet the label still read, in full, two words: Football. It made me sit down for a very long time.

This is the story of a meeting co-chaired by the Federal Minister for Economic Affairs, Ahad Cheema, and the Federal Minister for Power Division, Awais Ahmad Khan Leghari. The two ministers discussed a plan to conduct a technical audit of all Pakistani distribution companies (DISCOs), aiming to pin down two kinds of losses: power theft and technical and commercial (T&C) losses. Alongside that sits circular debt — the self-reinforcing chain of unpaid dues between generators, distributors and the state — and a solarisation pilot at Pesco and Qesco, the two utilities with the highest loss rates in the system.

I read the report slowly, twice, three times. This is the work I have done for 35 years. And across all 21 information points the system extracted, not one had anything to do with football. Not one. Every fact revolved around Terms of Reference (ToRs), an Expression of Interest (EoI), power tariffs, revenue shortfalls, and farmers in Punjab paying more for electricity than they should.

So why did a machine see football there?

I think I know the answer. And that answer cuts straight into my own profession.

When you train a machine-learning system to classify content, you do not teach it meaning. You teach it patterns. The word 'technical' appears heavily in sports copy — 'technical foul', 'technical area', 'technical director'. The word 'audit' has shown up in stories about club financial audits. The word 'committee' attaches to disciplinary panels and competition boards. The machine stitched the three pieces together, saw a high enough probability, and slapped on the label. It was not statistically wrong. It was semantically blind.

And that is exactly what is happening in modern football, only at a larger, more expensive, and far more dangerous scale.

I have watched this invasion since its early days, when analytics departments began sprouting across England like mushrooms after rain. In 2026, a Premier League club hired a PhD physicist as head of analysis. In 2026, a national team brought an entire data-engineering team into its training camp. In 2026, when I made my controversial prediction about Mohamed Salah breaking Luis Suarez's 31-goal record, I was mocked mercilessly. But my confidence did not come from watching football better than anyone else; it came from reading the metrics that analytics departments then undervalued: xG, burst speed, and Jürgen Klopp's pressing system. Salah scored 32 that season and won the Golden Boot. People called me mad. But my madness has its own logic.

I tell that story not to boast. I tell it to raise a larger doubt.

If a data-driven prediction can be right, then a prediction driven by misfiring pattern-recognition can also be right — in a dangerous way. The machine that labelled an electricity audit as 'football' is not a software bug. It is a scaled-down image, literally, of how football now operates: data analysts have slipped into the dressing room, carrying their spreadsheets, and started reading the game through columns of numbers with no breath in them.

I have sat in those rooms. I have seen a manager persuaded that a defender with elite xG-prevented numbers was the right signing, only to discover three months later that the player could not handle the tempo of a derby. I have seen a striker dropped because his expected-goals figure fell below par, while my eyes saw him opening spaces no metric captures. I have seen a club break its transfer record for a player based entirely on a fingerprint model, only for him to struggle all season because the team did not play the system the model assumed.

Every time, I think of that Pakistani electricity audit. A system sees a pattern, applies a label, and trusts the label absolutely. Sound familiar?

That is why I question the value of the metrics we now worship: PPDA, xG, xGA, progressive passes, packing rate. They are not wrong. They are merely severed from the actual rhythm of the match, just as a T&C loss ratio is severed from the life of a rural household in Punjab. The number is right. But nobody reads the story behind it.

And here is the point I want to make clearly: the biggest blind spot of modern football is not a shortage of data, but too many people reading data without understanding football.

I am not against data. I used data to predict Salah. I use data on air every week. But there is a fundamental difference between an analyst who understands football and one who only knows how to read a spreadsheet. The first looks at a team's PPDA and sees high-intensity pressure degrading because of a congested schedule — he immediately checks the fixture list, checks the holding midfielder's minutes, and trusts his eyes. The second looks at the same number and recommends sacking the manager.

Both are 'data-driven'. Only one reads the story correctly.

The machine that mislabelled the electricity audit is the embodiment of the second man. It sees the pattern. It does not see the meaning. And it never blushes when it is caught.

Trust me, I have blushed. Every mistake in front of a camera is a chance to rewrite your own story. When I mispronounced Modrić's name, I was forced to become a more careful commentator. When English viewers criticised me, I learned that credibility is built on the smallest details. From 2026 to now, I have spent twenty per cent of my writing time on verification: phonetic spellings, cross-checked figures, confirmed dates, sourced numbers.

The machine does not do that. The machine applies a label and goes to sleep.

But before you think I am telling a technology tragedy, let me turn a corner. Because I am not entirely sure I am right.

There is another possibility, one that forces me to be careful in judging that machine. Perhaps the algorithm did not mislabel. Perhaps the humans who designed the labelling criteria were the ones at fault. A machine-learning system does not invent the definition of 'football'. Someone wrote that definition. Someone chose the keywords, the signals, the probability thresholds. And if the person who wrote that definition has never sat through a match at three in the morning, never wondered why a centre-back chose to jump rather than drop off, then the definition will be as blind as the machine.

I wonder whether I am blaming the mirror instead of the face in front of it.

And there is one more thing that makes me hesitate. Throughout my career, I have used data to say things my eyes could not prove. I flagged Liverpool's pressing trend before other commentators saw it. I flagged signs of a team's physical decline before the table reflected it. Without data, those observations would have been mere feeling. Data gave that feeling a skeleton.

So why do I fear it?

Perhaps because I fear something very specific: I fear that one day a machine will label 'football' on a match I no longer recognise. A match optimised for metrics, where players run to a model, managers pick line-ups by algorithm, and the stands watch numbers move on a screen instead of watching people.

That is a real scenario. And it is not far in the future.

I have seen the signs in youth football. I have seen under-15 coaches asked to submit data reports to their boards after every match. I have seen fourteen-year-olds tracking their own biometrics on phone apps before they have learned to read a game by feel. I have seen academies teach kids to optimise personal xG before teaching them how to pass with the inside of the foot.

The Pakistani electricity audit is a small but concrete warning. The heart of football does not live in the stands; it lives in the sighs of those who stay behind. And if we keep reading football through columns of numbers cut off from those sighs, we will soon lose the very thing that makes football football.

What I take from the mislabelled audit is not 'data is bad'. What I take is a question about responsibility. Who is accountable when a machine mislabels? Who is accountable when an algorithm picks the wrong player? Who is accountable when a model misjudges a fine manager?

If the answer is 'nobody', then we are building a football in which no mistake can ever be corrected. And a football that cannot correct its mistakes is a football that has already died.

I write these lines after ten years hosting and producing 'Football Night', after early-career years in Madrid writing for Football Weekly and World Sports Daily, after being named SJA Sports Journalist of the Year (the John Bromley Trophy) five times, plus six other SJA awards. I do not say this to show off. I say it to affirm that I have walked enough forks in the road to believe data is a companion, not a teacher.

And I believe one clear thing: football needs translators, not worshippers. It needs people who can stand between a spreadsheet and a match, read both, and tell the rest of us the story that lies in between.

Fifty-one has taught me that impatience is a catalyst, but only when distilled through experience. I was impatient with the Salah prediction. I was impatient when I mispronounced Modrić. I have been impatient with many judgements across more than thirty years. But each time, I came back to an old question: what do my eyes see, what does my data say, and which of the two is telling the more honest story.

The machine that mislabelled the Pakistani electricity audit never asked itself that question.

I make one request of those running analytics departments at clubs and federations around the world. Let a former player, a former coach, or a 51-year-old former commentator like me sit beside your engineers. Not to deny the numbers. But to remind them that behind every number is a windy afternoon, a player with a sore knee, a stand waiting in silence.

If an electricity audit can be labelled 'football', the same can absolutely happen to one of your matches. That is what I believe. And that is what I stake my name on, just as I staked it on Salah in 2026.

I could be wrong. But as always, I am ready to answer for what I say.

And you, my reader, when you next look at a metric and feel it is too neat, too round, too clear to be wrong — I hope you remember this week's story. Somewhere a machine has just labelled a report on electricity theft as 'football'. And nobody in that system blushed.

That worries me more than a wrong prediction ever could.

I am still here, in Liverpool, writing on. And I still get the names right — of every player, every coach, every minister — even if people call me mad. My madness has its own logic. And that logic has not betrayed me yet.