Data Doesn't Lie, But the People Reading It Do: Lessons From a Fourteen-Page Report With Zero Numbers

On my desk in Valencia there is a fourteen-page document. I opened it at two in the morning, four hours after the match had ended and long after the event report had gone out. Page one had a title. Page two had a subtitle. Page five had tables. By page fourteen I finally noticed what I should have seen on page one: there was not a single number in it. Not a passing figure, not a heat map, not one expected-goals value. Every data cell had a label, and every data cell was empty.

People tend to believe the most dangerous thing in football analysis is a wrong number. I believed that for years. After holding a report generated from empty data, I understood something else: the more dangerous thing is a report that looks complete. A wrong number invites suspicion. A fully formatted report, with titles, tables and a table of contents, invites trust. And when someone has to fill the gap to hit a deadline, the output stops being analysis. It becomes a different product with no label on it.

That story is not only mine, and it is not only Valencia's. It is the story of an entire industry running on data pipelines that almost nobody inspects, only the water coming out of the tap.

Context: when football hires auditors for itself

Over the past fifteen years, European football has shifted from a game read by the eye to a game read through three stacked data layers. The first is raw event data, collected by providers such as StatsBomb, Opta and Wyscout, logging every pass, duel and player position to the hundredth of a second. The second is positional data from camera systems and in-ball chips, allowing the shape of a defensive block to be reconstructed in real time. The third is inference, where models turn raw data into something a coaching staff can act on: expected goals, pressure indices, progression metrics.

In Vietnam, these three layers did not arrive together. In recent years V.League 1 has begun to have its own event-data providers, and larger clubs have assembled small analysis departments, often just one or two people who handle both video and numbers. That is the right direction, but it creates a new pressure few discuss: when you have an analysis department, you must give it a deliverable. And when the deliverable is due on a schedule rather than on data quality, the content inside starts being written to fit the schedule, not the pitch.

I lived inside that loop at Valencia CF. In 2026 I joined the coaching staff as a tactical analysis assistant at twenty-four. In my first press room, an older male reporter asked whether I really understood Marcelino's pressing block or was only there for decoration. I said nothing. That weekend I sent the staff a fourteen-page report showing the team lost sixty-two percent of ball control on the left flank, forcing Marcelino to adjust his line-up for three straight matches. Nobody questioned my gender again.

But what I learned was not the power of numbers. What I learned is that numbers are only powerful when they exist. Most of the time, the hardest part of this job is not analysing the number. It is confirming the number was actually measured.

Core analysis: anatomy of an empty report

When the input data is empty, the report template does not collapse. It fills itself in. That is the biggest blind spot in modern football analysis.

Picture a typical data pipeline at a professional club. After the match, the event-collection system runs and pushes the raw file to the server. The next morning, analysis software reads the file and generates a report of dozens of pages: possession duration, passes by zone, average position maps, pressure per defensive action. The analyst opens it, adds written judgement, sends it to the staff. The staff read the judgement, nod, adjust next week's training.

In that chain, one link is almost never inspected: the first step. If data collection fails, and fails silently, every later step proceeds normally. The software raises no error, because it only reads the file; it cannot know whether an empty file means a fault or a match with nothing to record. Empty cells are treated as zero or skipped, and the report still appears with full headings. The pages still have a table of contents. The tables still have their lines. Only the numbers are missing.

This is where the profession splits in two. The first group sees the empty cell and stops, calls the technical department, says we have a problem. The second group sees the empty cell and writes on, because the report is due, and because a story about the match is already in their head from watching it live. They replace the number with memory. And memory is never empty, even when it is wrong.

I was once in the second group. Not out of a wish to deceive, but because delivery pressure is very real. When your report is due at seven in the morning and you know the team lost the ball on the left because you watched it, writing an approximate number feels harmless. It is not. That approximate number will be used to explain a tactical decision, and that decision will affect a real player, on a real day, on a real pitch.

Data doesn't lie, but the people reading it do. And the reader most likely to lie is the one forced by a deadline to reach a conclusion.

Three common substitutes

The first is the memory substitute. The analyst watches the match, remembers a few standout moments, and when data is missing writes from that memory and presents it as measurement. It is dangerous because it blends two different things: a subjective impression and an objective reading. The reader cannot tell them apart, and so trusts.

The second is the template substitute. The analyst takes last match's figures, or the season average, and assigns them to this match. It looks highly professional. The number is real; it just belongs to another game.

The third is the model substitute. The analyst reruns a prediction model on incomplete data, the model returns a mean, and that mean is printed as a result. This is the most dangerous because it looks the most scientific. A model run on missing data does not return truth; it returns the assumptions of whoever built it, dressed in arithmetic.

All three lead to the same place: a professional report about a match that was never measured.

Volgograd, and the heat that was not in the model

I learned this at the 2026 World Cup, in England against Tunisia in Volgograd. Before the match I predicted England would press high in the Guardiola manner, based on their qualifying pressing data. I was wrong. That afternoon's temperature reached thirty-four degrees Celsius. England's players ran an average of 9.2 kilometres, 1.8 kilometres fewer than in their previous match. They dropped the tempo, Tunisia produced five dangerous shots, and Gareth Southgate said afterwards that he had deliberately reduced intensity because of the heat.

What matters is not that I mispredicted. Everyone mispredicts. What matters is that I analysed tactics on a sheet of paper with no temperature on it. My data was technically complete but environmentally incomplete. And that gap did not appear as an empty cell. It appeared as a finished conclusion.

After that match I built an attached data sheet for every game: weather, humidity, pitch type, travel time, days of rest, even the tendencies of the appointed referee. That sheet is longer than the tactical one. It makes my conclusions less confident, and that is a good thing.

xG, PPDA, and the trap of a correct number in the wrong context

Expected goals, the most cited metric in modern football, is a perfect example. An xG value is calculated from shot location, shot type and the situation leading to the shot. A shot from inside the box is worth more than one from outside. That is true. But if positional data is skewed, or if the model was trained on a different league than the one being analysed, the value is still printed to two decimal places, and readers still believe it.

I once read an internal club report in which a striker's xG spiked over five consecutive matches. The staff concluded the player was in form. In reality, his shot count fell while shot quality rose, because three opponents in that run were weak teams. The number was right. The context was wrong. The conclusion was wrong.

That is why I always check pressure per defensive action, PPDA, alongside match context, and never read PPDA alone. A low PPDA can mean a team presses well, or that the opponent passes too much in their own half, or that the team is losing and forced to push up. Three causes, one number, three different conclusions. Without context the number means nothing. It only appears to mean something.

From the data pipeline to the betting market

One thing this industry tends to avoid. Raw event data, in its most real-time form, does not only flow into a club's analysis room. It flows into a parallel pipeline where betting companies buy access. Live data supplied to betting companies is the darkest side effect of the digitisation of sport.

This connects directly to the empty-data problem. When the same source feeds both an analysis room and an odds board, a fault at the collection step does not merely produce a bad report. It produces a false market signal. And in a betting market, a false signal can be exploited by those who know it is false, before anyone at the club notices the empty cell.

I do not write this as an accusation. I write it as an audit line item. If a professional club builds its decision process on data bought from a provider, it needs to know how many hands that data passes through before it reaches its own meeting room. Auditing part of a pipeline is not auditing.

A stalled season and a notebook of nine matches

In 2026 the pandemic stopped Spanish football for three months. Valencia CF slid into financial crisis, could not pay wages, and the press filled with takeover rumours. As a member of the coaching staff, I was the only one who stayed in contact with the players by video. Rather than panic, I did what I know: I took inventory.

I built a notebook analysing the last nine matches based on pre-pandemic data. I had no data for the break, because no matches were played. I knew that, and I wrote it on page one: this is a pre-pandemic inventory, any comparison with the later period is flagged inference, not measurement. When the 2026-21 season opened, my articles on Valencia having to shift from a four-four-two to a three-five-two for lack of strikers were republished by a major football site.

What I learned was not that my notebook was good. It was that stating clearly which parts are data and which are notes matters more than the notebook's content, because whoever cites it later will know exactly what they are citing.

V.League, and the lesson not yet written

I follow Vietnamese football from a distance, long enough to see a familiar paradox. V.League 1 has huge stadium audiences and vast online debate, yet public data infrastructure remains thin. Most tactical argument here happens by eye and by feeling, and that is not wrong. What is wrong is when it wears the clothes of data.

I have read many analyses of Vietnamese clubs citing very specific figures with no source. Passes, pressing counts, kilometres covered. Those figures may be right. But when I traced them, most could not say which provider produced them, by what method, on what sample.

This problem is not Vietnam's alone. But it has particular consequences in a league building its analytical foundation from scratch. If the first analytical generation in V.League learns the habit of using numbers to lend authority to a conclusion it already held, the next generation will treat that as the standard. A wrong standard is harder to fix than a wrong report.

Contrarian angle: the execution blind spot is not the players

When a team loses, the public reflex is to find fault with players or the manager. They are visible, named, on television. The real execution blind spot is somewhere far less visible: at the data-collection step, at the gatekeeper who checks data, and at the deadline for the report.

Before asking why we lost, ask what we prepared for. And before asking what we prepared for, ask whether we have the data to know what we prepared. Those three questions are usually collapsed into one, and the third is almost always skipped.

I have seen major personnel decisions made on a report whose data section only its author knew was empty. I have seen players undervalued because their numbers were low in a match where the positional tracking system failed. I have seen two-hour tactical meetings arguing over a figure that turned out to be a software default value for missing data.

A rule written in blood, not ink. The rule I set for myself after those episodes is simple: no report leaves my desk without a line stating the data source, the measurement method, and the confidence level of each metric. If a metric has no source, it comes out. If it comes out, I must write in words that I have no data for that part.

This makes my reports shorter, and sometimes makes me look indecisive. I accept that. An honest short report beats a long wrong one.

But I must also be honest about the other side, the one my peers rarely mention. Constantly flagging confidence levels can become a shield. When every conclusion is framed in probability, no conclusion can really be challenged, and the author never has to answer for a wrong call. That is not science. It is evasion decorated with jargon.

So at the end of every report I write, I force myself to give one clear conclusion, no probability, no conditions. I may be wrong. But if I am wrong, I am wrong in a way that can be checked, and next time I correct it. A wrong call that can be corrected is better than a vague call that is never wrong and never right.

What is genuinely different next match

After all this, the question I ask myself before each new match is no longer how this team plays. It is what in this match is genuinely new, and whether I have the data to prove it is new. Precedent is a good teacher but not the only one. A phenomenon that has never occurred has no precedent, and if I only search the past I will miss it.

In Vietnam, the new thing may be a V.League club having an analysis department large enough to split data collection from analysis. That is a small but correct step, because it creates a second person who can see the empty cell the first one missed. In auditing this is called separation of duties. In football it is often dismissed as wasted headcount. I think the opposite.

When you have a data pipeline serving decisions about people's careers, having one person whose only job is to check whether the pipeline is still flowing is not waste. It is the minimum condition for calling yourself professional.

Takeaway

The press room is not for the timid, it is for those with numbers. I would add a clause that years in this trade taught me: the press room is for those with real numbers, and with enough courage to say when they have none at all. An honest empty cell can always be filled by one correct measurement. A silently invented number stays in the system a long time, gets cited, gets repeated, and eventually becomes part of a history that never happened. The question for anyone writing about football, on any platform, is not how confident you are. It is whether you know where your number came from.