Inside Football's Data Supply Chain: When an Empty Input Still Produces a Conclusion

The meeting room on the eleventh floor smelled of cold coffee. In May 2026, fourteen people sat around a long table in Guangzhou, and none of them talked about football. They talked about payment clauses. The Premier League had been suspended since March, La Liga had stopped, Serie A had stopped, the Champions League had been postponed with no return date. But the broadcasting contract kept running on its own calendar: matches to be delivered, hours to be aired, money to be wired before the fifteenth. The product had ceased to exist. The machine kept producing output.

I sat at the end of the table taking notes. Someone proposed a three-month payment deferral. Someone proposed renegotiating the price. Someone proposed re-airing old matches and calling it special content. Nobody proposed the cheapest option of all: telling the audience plainly that we had nothing for them today. Throughout that meeting one thought kept returning to me. This industry is built to always have an output, whatever the input happens to be.

Three weeks later a friend who worked at a sports data company sent me an internal scouting report. He called it clean. It had every column: top speed, passes per ninety minutes, heat map, expected goals, comparison against positional peers. There was just one small detail: the league it analysed had not played a single minute in ten weeks. The numbers were still there, tidy, confident, with not one question mark.

That was when I started caring about a question that has since become the axis of almost everything I write: when the input is empty, what produced the output?

Four layers, and the layer nobody audits

The football data supply chain has four distinct layers. The first is capture. In Europe's major leagues this is handled by camera and sensor systems: Hawk-Eye in the Premier League, Second Spectrum in several European and North American competitions, optical tracking rigs around the pitch. A single match generates millions of positional data points. In Vietnam's national championship, most of this work is still done by a person in the stands tapping events by hand, or by outsource firms logging from video files.

The second layer is aggregation. Companies such as Stats Perform with its Opta product, Sportradar and Deltatre collect raw data, standardise event codes, label them and redistribute to clients.

The third layer is cleaning. This is the least discussed and the most decisive layer of the whole chain. Data cleaners check every anomalous event, fix wrong coordinates, cross-reference video, close tickets. Most of them sit in low-labour-cost locations. A significant share of them are women.

The fourth layer is modelling and distribution: club analytics departments, bookmaker trading desks, broadcaster graphics teams, match-tracking apps, statistics sites.

Four layers. The first three have contracts, service-level commitments and penalty clauses. The fourth has consumers. And between the third and fourth there is a gap that has never once been audited: when data is missing, who decided what to put in its place?

Data does not lie, but the people who clean it sometimes do.

2026, and the lesson I thought I had finished learning

In 2026, in the AFC Champions League quarter-finals, I used positional data from twelve sensors around the pitch to prove something the naked eye could not see: the visiting side's 4-2-3-1 was actually becoming a 3-4-3 in possession, stretching the opposing back line horizontally and leaving a permanent gap in the right inside channel. Three days later that team's head coach confirmed exactly this in his press conference.

A male colleague in the newsroom had once told me that women only know how to read numbers, not how to understand football. After that day he never said it again. The analysis was shared eight thousand four hundred times. My under-25 audience grew by two hundred and ten per cent.

I thought I had won. In truth I had learned only half the lesson. The other half arrived in June 2026, from a stand in Nizhny Novgorod, when I mispronounced a Croatian player's name three times in the first half. Social media reacted within four minutes. The overconfidence from the previous year had made me skip the identity check.

That night I did not delete the clip. I rewatched the whole match and noted pronunciation from the source language. Over the following thirty days I built a standard pronunciation table for seven hundred and thirty-six players and published it free. A 736-name pronunciation table is not discipline; it is an apology that has been systematised.

Those two stories, placed side by side, show how the data supply chain actually operates. 2026 taught me that data can see what the eye misses. 2026 taught me that data is only as right as the person who prepared it.

Anatomy of an empty input

Back to the central question. When a football event has no coordinates, no timestamp, or does not exist at all, what happens inside the pipeline?

The short answer: the pipeline rarely stops. It fills the gap.

In statistics this is called imputation. A missing coordinate is replaced with the average coordinate of similar events. A minute of lost signal is interpolated from the minute before and the minute after. A missing metric is assigned the median value of positional peers. Technically these are reasonable choices. Cognitively, this is where the truth starts bending without anyone signing their name to it.

The problem is that imputed values carry no label. When you open a statistics sheet, you do not see behind the figure 8.4 a sentence reading: this cell was imputed. You see 8.4. And you believe it.

The same happens with expected-goals models. Such a model is trained on hundreds of thousands of shots with complete data on location, angle, body part, defender density and the type of pass that arrived. Fed a shot with missing data, the model does not return a blank. It returns a number, drawn from learned defaults. The consequence is that every match has metrics. Including the match where the capture system failed for half a half.

I once cross-checked a Vietnamese domestic league match whose positional data had been lost for almost the first forty minutes of the second half. The post-match statistics sheet still displayed fully. No note. No warning. Fans read that sheet, saw the home team had completed fewer passes, and concluded the home team had been pinned back. In reality, most of those passes were simply never recorded.

Core insight: the modern football data system is engineered to always produce an output, not to always have an input. In that architecture, a gap never becomes a gap. It becomes a number.

This is the point I most want Vietnamese fans to grasp, because we consume a great deal of imported data we have no way of verifying. You watch a Premier League match at three in the morning, graphics appear with this metric and that metric, and you have no way of knowing whether the figure came from a sensor or from imputation. Both sources look identical on screen.

Who cleans the data, and under what conditions

There is one side of the supply chain that analysts rarely discuss, even though it determines the reliability of everything: labour conditions at the cleaning layer.

Football data cleaners work by ticket. Each ticket is a suspected event. They open the video, find the exact frame, determine the event type, assign coordinates, close the ticket, move to the next. Targets are usually measured in tickets per hour. In many cases bonuses only appear above target.

You can see the problem. A system that pays per closed ticket rewards the speed of closing, not the accuracy of what is closed. The cleaner has no incentive to write in the notes field that they are unsure. Writing unsure means reopening, losing time, missing target.

So the nominally cleanest layer is the layer under the greatest time pressure. And that layer almost never appears in any audit. Nobody audits the cleaner. People only audit the number the cleaner produced.

This is why I always ask one question before using any statistics sheet, including sheets I commission myself: who handled this, for how long, and what were they paid to do?

The transfer market: the perfect empty-input machine

If you want a market where output always exists regardless of input, look at transfers.

Every window produces thousands of rumours. Most have no source. Some have a source, but the source is an agent, meaning a party with a direct interest. A smaller share comes from clubs, usually to move a price. In every case the final product reaching fans is identical: a name, a number, and a verb in between.

Every source has a motive, and every motive has a price. The agent wants leverage for a new contract. The club wants to inflate a player it is about to sell. The team wants to reassure its supporters that it is acting. Some outlets want traffic. None of them is lying in the legal sense. They simply feed an empty input into the system, and the system does the rest.

In June 2026, in Bucharest, France were eliminated in the round of sixteen on penalties. Kylian Mbappe missed the decisive kick. Amid the European pile-on, a friend in the transfer world told me a detail that appeared in no bulletin: Real Madrid had submitted a one hundred and eighty million euro offer for Mbappe, and Paris Saint-Germain had rejected it. The twenty-two-year-old walked into that match with his mental state already broken before the ball rolled.

I wrote a three-thousand-word piece. I did not defend him. I explained the psychological mechanism of a human being converted into a number, and what happens inside him when that number is refused. A French newspaper cited it.

Data only becomes rebellious when someone is brave enough to believe it.

The professional lesson here is concrete: never write about a single match without placing it in the economic and transfer-market context around it. A missed penalty in the eighty-eighth minute has less to do with shooting technique than with the bank balance of a club two thousand kilometres away.

Four times football paid the price for an empty input

Transfers are only one market. Higher up, the same architectural fault has repeated several times, at far greater cost.

In December 2026, the Mediapro group lost its Ligue 1 broadcast rights after failing to pay a contract worth eight hundred and fourteen million euros per season. French football had sold a product on the assumption that the buyer would always have money. When that assumption collapsed, the league had to resell at roughly a third of the price. By the 2026-25 season, Ligue 1's total domestic rights value had still not recovered its old level.

In February 2026, the Premier League charged Manchester City with one hundred and fifteen financial rule breaches. In November 2026, Everton were docked ten points, reduced to six on appeal. In March 2026, Nottingham Forest were docked four points. Those three events, placed side by side, show something simple: football has control thresholds at the financial layer, gates at the player-registration layer, conditions at the eligibility layer. Only the information layer has no gate at all.

In December 2026, the Premier League announced a domestic rights package for 2026-2029 worth six point seven billion pounds. That figure was not priced on the quality of accompanying analysis. It was priced on the scarcity of premium live content. Money flows into the input layer, and nobody pays for the verification layer.

Minimum content threshold

From all of this I draw one principle I apply to myself, and which I believe the industry should apply systemically.

It is called a minimum content threshold. Before any process is allowed to generate a conclusion, it must clear a gate: at least one named entity, and at least one verifiable fact. Otherwise the required result is a blank line, not a sheet full of numbers.

This sounds obvious. In practice almost no football system applies such a gate. Analytical pipelines are designed to run to completion regardless of incoming data. When the extraction step returns an empty list, the downstream analytical step still fires. The fault is not in the output. The fault is that nobody halted the chain.

I built my own gate by hand. That is why the 736-name pronunciation table exists. It forced me to look up every name, pronounce every name, record the source, cross-check before going on air. No software did it for me. But that manual process was what revealed that some names I had assumed were English were originally Portuguese, and that certain vowels I had mispronounced for twenty years failed not because I was bad but because nobody had ever corrected me.

My most valuable mistakes exist in 736 versions, and all of them were worth making again.

Vietnamese football inside that supply chain

One reality I have observed across many years working between two football cultures.

Vietnamese football is a consumer of data, not yet a producer of it. We import statistics sheets from the Premier League, La Liga and the Champions League, and we consume them faster than we verify them. When a Vietnamese player is first valued by a foreign site, within twelve hours that number has become fact in online arguments.

At the capture layer, most domestic league data is still logged by hand. That is not inherently bad, provided everyone knows it is logged by hand. The problem arises when hand-logged data is presented exactly like sensor data.

At the youth development layer, the story is more complicated still. Scouting networks in developing countries have a dual character that few are willing to state plainly. The same network that discovers genuine talent also produces what I call football lottery tickets and broken families. When a fifteen-year-old is taken abroad on the promise of an academy place, nobody records the success rate. Nobody cleans that data. No statistics sheet displays the families left behind.

That is the biggest gap in the supply chain we consume: inputs that are never collected, and therefore never become outputs.

The stadium with no singing

There is one kind of empty input I care about especially, and it does not live in a computer.

In May 2026, when European football returned after the pandemic, stadiums opened without crowds. I watched a match on screen and realised the first thing lost was the singing. Not the noise. The singing. Noise can be patched with artificial sound. Singing cannot, because it is a structured signal, reacting to the flow of the match, and it carries information.

A stadium with no singing is a disconnected sensor. The statistics sheets were still full. The metrics still displayed. But an important data layer had just vanished, and nobody halted the process.

In a stadium with no singing, I heard the future of broadcasting.

The old fan model relied on physical presence. You went to the ground, you sang, you generated data nobody recorded but everybody felt. When the pandemic forced everyone home, part of that model broke, and I realised we had overlooked an enormous interaction channel. Supporters never stopped wanting to talk about football. They simply stopped having a place to talk.

That is why I launched a live stream in May 2026, analysing the 2026 Istanbul final between Liverpool and AC Milan, inviting viewers to propose tactical changes minute by minute. Management where I worked refused to fund it, arguing that audiences only want live matches. The show reached two hundred and fifty thousand views, fifteen times a second-tier commentary match.

Fans do not leave the stadium when they bring the whole stadium into their living room.

The contrarian angle: what is missing is not good data

Here I want to bet on a position I know will be uncomfortable.

The popular explanation for every analytical failure in football is bad data. It sounds reasonable, and it is useless. Because if the only problem were bad data, we would simply buy better data. Yet the biggest broadcasters, the richest clubs and the leading data companies have all bought the best data available, and they still commit the same error.

The problem sits in the business model, not in data quality. The sports media industry pays for output, not for accuracy. A piece of writing or a graphic that does not exist generates no revenue, no views, no advertising. A piece of writing with errors, published on time, generates all three.

This leads to a counterintuitive conclusion: the most honest answer in this entire industry, the sentence I do not know, is the only sentence nobody is paid to hear. And so it is almost never spoken.

I once thought a twenty-four-hour correction culture was an ethical standard. I now think differently. It is a survival compromise. We accept error first and fix it later because the system does not let us wait. The data cleaner is at fault, but the person who designed a system paying per closed ticket is the one who created that fault.

And there is something harder to hear: sometimes an empty cell is the single most important piece of information in the whole table. A metric that does not exist in a domestic league tells me more about that football culture than a beautiful imputed metric. The silence of a stand tells me more than a full one. We are trained to read what is present, when our job should be to read what is absent as well.

What I want fans to do after reading this

I am not writing this to tell you to stop trusting data. The opposite. I am writing so that you trust data conditionally.

Concretely, next time you watch a match and a metric appears, ask yourself three questions. How was this collected. Who cleaned it. And if it was missing, what was put in its place.

Those three questions require no technical knowledge. They require only the habit of doubting in the right place.

For Vietnamese football, I believe this is the right moment to start building our own verification layer instead of only consuming other people's output layer. A domestic league dataset with clearly recorded sources, methods and known gaps is worth more than ten flawless datasets with no provenance. And that does not require a large budget. It requires a decision.

The question I leave for you, and for myself in the next piece, is this. If fans one day began paying for honesty instead of speed, how much would this industry change within twelve months.

I do not know the answer. But for the first time in thirty-nine years of working in this trade, I am glad that I do not.