When Football Data Mislabels: Lessons from a Pakistani Power-Sector Audit

There is a file in my system that forces me to rewind every time I open it. It carries the label "Football." Reading it from the first line to the last, there is not a single team, player, coach or match anywhere inside. What it contains is a technical audit of Pakistan's electricity distribution companies, the group known as DISCOs, co-chaired by the Federal Minister for Economic Affairs, Ahad Cheema, and the Federal Minister for Power, Awais Ahmad Khan Leghari. Twenty-one information points in the original record, and all twenty-one revolve around electricity: technical and commercial losses, circular debt, wrongly issued bills. Not one of them touches football.

Someone has just blown the whistle incorrectly. And as on every occasion I have sat in the VAR room, I know the culprit is rarely a villain, but a system nobody ever reviewed.

Content tagging is the least glamorous job in a newsroom. Nobody writes articles about it, nobody argues about it. But it is exactly like an assistant referee: one flag raised wrongly, and the entire move behind it goes off course. In this case, the algorithm saw the word "technical," saw the word "companies," and applied the label "Football." One word. One tag. It sounds small. But a mislabelled file will drift into exactly the pipeline it does not belong to, and there it will be read, learned and amplified.

I have been in a similar position before, and the price was not small. In November 2026, at San Siro, Milan hosted Juventus on matchday 12 of Serie A. I was the assistant VAR in the operations room. In the 56th minute, Higuaín scored to make it 2-0. The feed showed the Argentine striker was offside by 0.2 metres. I hesitated, afraid of being wrong, and I did not recommend a review. Milan lost 0-2. After the match, the referee supervisor criticised me in front of the whole team. I spent a month reviewing 47 similar incidents and built a 37-point checklist to standardise every decision. Since then, I have never issued a judgment without a foundation.

Before I blow the whistle, I review myself. That is why I am writing this.

The original report describes a clear administrative process. A committee co-chaired by two ministries was set up to review the entire power-distribution system. The goal: identify electricity theft and quantify both technical and commercial losses. The instrument: an independent technical audit, beginning with an expression of interest, an EoI, and a set of terms of reference, the ToRs, expected to be finalised the following week. The IT ministry was tasked with technical support. Two entities were named specifically, Pesco and Qesco, the highest-loss, highest-theft distribution companies. In parallel, solarisation pilots were being deployed to ease the grid. And behind all of it sits circular debt, the self-reinforcing chain of unpaid dues between generators, distributors and the state.

This is a report about infrastructure governance. It has its own value. But it contains not one word that belongs to football.

What is worth noting is that the "Football" label did not come from a saboteur. It came from a systemic blind spot. In the data industry, this is called domain misclassification, and it is the most dangerous kind of error, because it is silent. A collision on the pitch makes the stands scream. A wrong tag makes nobody scream. It simply sits there, waiting to be used.

The core analysis begins with a simple question: if a file about a power audit can slip into a football database, how many mislabelled files are already sitting in that football database without anyone knowing?

I do not have that number. But I hold one professional conviction: every analytical model is only as trustworthy as its input layer, and its input layer is only as trustworthy as the mechanism that checks it. A power-sector audit labelled as football damages no match. But it reveals that one stage of the pipeline is running without anyone reading it back.

I do not trust my eyes; I trust the slow-motion replay. In football, we are used to the idea that a referee's decision is provisional until VAR has finished. We accept that the human eye can deceive. Yet when it comes to data, we treat the first label as a final ruling. That is a contradiction, and it does not stay contained inside one newsroom.

My 37-point checklist, built after that night at San Siro, has no entry for trusting a label. The first item is always: where did this data come from. The second: who verified it. The third: if it is wrong, what happens. A power-audit report labelled as football would fail item one immediately, and it should have been stopped right there, not discovered only at the deep-analysis stage.

What is interesting is that the evaluation process corrected itself. When this report was put through deep analysis, the first conclusion was not an opinion about football, but a refusal: the football framework does not apply to this document. In principle, that is a perfect VAR intervention. The flag went up the moment the frame showed something abnormal, and nobody bent the frame to fit a pre-written conclusion.

But look at the verification side. All twenty-one information points in the record carry an empty source field. No citation, no date, no issuing authority. Technically, that is a goal with no available camera angle. For a referee, it is the worst condition: you are forced to decide without enough data to decide.

The consequences fall into three layers. The first is the domain-misclassification error, rated high, and the fix is to remove the file from the football dataset before it can contaminate anything. The second is the missing-source problem, rated medium, requiring independent verification of the named individuals and stated decisions. The third is its announcement-stage nature, since the audit is still an intention, not an outcome. All three layers belong to energy, not football.

On governance, football has its own rulebooks: financial fair play, transfer registration, disciplinary sanctions, competition eligibility. Not one of them touches this document. The only thing here that can be called law is the power-sector regulatory framework, where an inter-ministerial committee sits and decides that a transparent audit is needed. The difference is not in form but in subject. One regulates how clubs spend money. The other regulates how a country distributes electricity.

The pressure in this material is not sporting pressure either. It is political and public-service pressure, coming from consumers who were overbilled and from areas suffering chronic blackouts. The stands here do not sing; they protest. The coaching staff here are not sacked; they are questioned in parliament. Both are pressure, but they belong to two entirely different frames of reference, with no conversion rate between them.

The solarisation rollout at Pesco and Qesco is a notable strategic design choice. It resembles a club deciding to decentralise its scouting network instead of concentrating everything in a single data centre: placing generation capacity close to consumption and cutting off the bottlenecks most easily exploited. But as with any restructuring, success depends on execution, not on the press release.

Then comes the operational dimension. The audit depends on an expression of interest, on expert recruitment, and on terms of reference that are not yet finalised. In other words, it depends on whether people actually do what they have announced. In football, we call that the gap between the pre-match statement and the performance after the opening whistle.

And here is where I see the real link to my own trade. Data analysts are pouring into dressing rooms, press rooms, and transfer contracts. They arrive with elegant models, smooth charts, indicators nobody dared to mention a decade ago. But many of those conclusions rest on data layers that were never checked, exactly like that wrong label. The conclusion is tidy. The real rhythm of the match is not.

This is where I want to be a little counter-intuitive. Most people's first reaction is to blame the algorithm. I think that is a misreading. The algorithm did exactly what it was taught: it read keywords, it assigned a tag, it passed it along. It has no concept of truth, and it does not need one. The fault lies in the absence of anyone at the final check, and in our assumption that a correct label means a verified label.

The real danger is not the file that is obviously mislabelled. The danger is the mislabelled file that still reads plausibly. A document that is ninety percent correct and ten percent fabricated will slip through every superficial check, and then be used to predict, to price, to decide. An obvious error is easy to fix. A hidden error stays. It stays longer than the career of the person who made it.

Technology does not kill football; it kills blind faith. The death does not come from machines reading wrongly, but from us no longer reading back after the machine has read. And in an industry where every transfer decision, every contract, every starting spot can be shaped by a spreadsheet, ceasing to read back is a slow form of self-destruction.

Every ruling deserves one review, including the ruling of the data. Including my own.

So what should be done? I do not believe in a world without error. Football is a game of margins of error, and so is data. What I believe in is a checking habit built into the system, not a promise to be more careful tomorrow. A mandatory verification step before a file enters the database. A flag-raiser at the end of the pipeline, not at the start. And an attitude: reviewing is a duty, not a suspicion.

The question I want to leave behind is not how to make the algorithm never wrong, because that is impossible. The question is who will read it back, at what hour, and whether that person has the courage to raise the flag when everything on the screen still looks fine.