Full Template, Empty Data: Why a Null Result Is More Trustworthy Than a Beautiful Analysis

Full Template, Empty Data: Why a Null Result Is More Trustworthy Than a Beautiful Analysis

7:12 a.m., 9 July 2026, in a fourth-floor flat in the Ruzafa district of Valencia. I open the file the desk sent overnight. Inside is a Stage-1 extraction of an article about the transfer window, with an instruction attached: produce the Stage-2 deep analysis, all nine dimensions, all tables, full length.

The template has eleven fields. Two carry values. The "domain" field reads: football. The "article type" field reads: unclassified. The remaining nine are blank. The "information points" list — the thing the entire Stage-2 analysis is supposed to hang on — is an empty list. The "entities involved" field contains one instruction: identify from the information points above. There are no information points above.

I sit still for about four minutes. This is the kind of file the industry processes in exactly one way: fill it in. Insert a club name, a player name, a season, and let the template generate words. Nine dimensions and more than forty cells will automatically produce three thousand words that read as highly professional: tactical analysis, financial structure, opinion cycles, injury risk, transfer pathways. Coherent. Fluent. And untrue.

I take the other route.

Why an empty template is the best stress test there is

In twenty-eight years in this trade, I have never met a tool as useful and as dangerous as the analysis template. When it has data, it forces discipline on the writer: every claim must point back to an information point, every figure must have a source, every conclusion must survive the question "based on what?". When it has no data, it becomes a machine that manufactures the tone of expertise. That is the difference between a spreadsheet and a speech: an empty spreadsheet is empty, but an empty speech can still be delivered with great conviction.

The football industry now runs largely on that kind of speech. Every transfer window, the volume of rumours published on aggregation platforms exceeds the number of deals actually completed by a factor of many tens. Nobody pays for a rumour that is later disproved. But nobody is fined for publishing it either. The incentive structure sits on the side of volume, and the result is that readers drown in a current where the density of real signal keeps thinning out.

An academy is like an archaeological stratum: whatever layer is rushed, that layer collapses. That holds for football academies, and it holds for the trade of writing about football academies.

Readers come to me not because I write better. They come because they need a filter. They want to know which story has evidence and which story has only verbs. And the first filtering tool is not professional intuition — intuition is built later. The first filtering tool is an honest template, in which a blank cell is allowed to remain blank.

This morning's file is an honest template. It says plainly that it has nothing. The problem is that most of the content production line refuses to read that plain statement.

Triple verification: a real procedure, not a slogan

I verify every number three times before it reaches the page. Not out of aesthetic perfectionism, but because I have paid the price for not doing it.

The first pass is source verification. Who recorded this number, by what method, under what conditions? A passing-accuracy figure can differ by several percentage points purely because two data providers define a long pass differently. A distance-covered figure can differ by half a kilometre purely because the speed threshold used to define "high-intensity running" differs. Without a definition, a number is just an opinion wearing a lab coat.

The second pass is context verification. Competition level, game state, sample size, opponent, point in the season. A midfielder who keeps the ball well against a side already two goals down and no longer pressing tells you very little about his ability against a deep block. Stripping numbers from context is the fastest way to lie without telling a lie.

The third pass is cross-verification against an independent source, or against my own tracking log. My tracking log is a notebook and a spreadsheet, not a machine-learning system. It is slow. It is also the only thing whose method I control completely.

Three passes take time. In a trade that pays for speed, this is a genuine competitive disadvantage. But I chose it very early, and I do not intend to change.

I arrive at the stadium later than everyone else, because I have read the spreadsheet before I read the match.

Paterna, April 2026: when the data actually existed

In April 2026, I asked for access to the Paterna training ground to watch a friendly between Valencia's Juvenil A and Villarreal B. A few male reporters were in the press area that day, and their conversation was about who scored.

I stayed behind with a notebook. The player wearing number 7, born 29 February 2026, was Ferran Torres. My hand-kept log recorded nine successful dribbles, four chances created and one assist. I should be clear: those are numbers I counted myself, not figures from a data provider, and I always label them as such whenever I cite them. But the more important thing was not the number. It was the positional chart I redrew afterwards: he kept drifting inside rather than hugging the touchline. He received the ball in the inside channel, in the space between the opposition's midfield and defensive lines, not on the chalk.

I wrote a piece of roughly two thousand words for a digital technical magazine, arguing that this movement pattern suited an inside-forward role better than a pure touchline winger.

On 30 November 2026, Ferran Torres made his first-team debut for Valencia in a Copa del Rey match. What followed is widely known: a move to Manchester City in 2026 for a reported fee of about 23 million euros, then to Barcelona in January 2026 for a reported fee of about 55 million euros plus add-ons.

Every star was once a forgotten line of data.

I retell this not to congratulate myself. I retell it to draw a comparison. The 2026 piece existed because there were nine dribbles, four chances, one assist and a positional chart. It had raw material. The template I received this morning has no raw material at all. The difference between those two situations is the entire content of this trade.

Kazan, 15 June 2026: using numbers instead of arguing

At the Kazan Arena, for the 3-3 draw between Spain and Portugal, I was one of four women in the press room. I asked about the space behind Spain's midfield. A few people laughed.

That evening I stayed behind with my tracking log and split Portugal's three goals into three separate mechanisms, because lumping them together is the laziest form of analysis.

The first came from a first-half set piece — a designed routine, not a random bounce.

The second was the goal the press room would pin on goalkeeper David de Gea the next day. My log showed the passage started earlier: Sergio Busquets was dragged out of his position in a duel moments before, Spain's screen lost its structure, and Cristiano Ronaldo found the space to shoot from outside the box. The goalkeeper's technical error was the last link in the chain, not the first.

The third was an 88th-minute free kick. The mechanism there was different again: I recorded that the foul leading to the free kick began with Spain's midfield line being knocked out of shape by substitutions, which stretched the distances between the lines and made a late tackle almost inevitable.

I also noted that Portugal's defensive line sat at an average height of roughly 52 metres from goal for most of the match — a deep block that ceded territory and lived on transitions. That was a deliberate choice, not passivity.

The analysis ran the next day, laying out those three mechanisms. The following day, head coach Fernando Santos referred to the piece in his press conference. A national radio station called to invite me on as a commentator.

What I learned in Kazan was not that I was right. What I learned was that numbers do not argue loudly; they simply stand there and force others to respond. To work that way, the first thing that must exist is the number. Without numbers there is only voice, and in a press room, the voice of a then-36-year-old woman carried very little weight against laughter.

Knees and fear: where medical data is abandoned longest

There is one area where sloppy analysis does the most damage, and almost nobody reads it carefully: anterior cruciate ligament injuries.

Cohort studies published in sports medicine put the re-injury rate for ACL-injured athletes returning to pivoting, change-of-direction sports in the range of 15 to 25 percent, with the higher rates concentrated among those under 25. I say range, because studies differ on how they define re-injury, on follow-up periods and on inclusion criteria. But that band is enough to state something professional football routinely ignores.

The problem is not the surgery. The problem is the return schedule.

A 22-year-old winger returning at month eight instead of month eleven generates three different decisions inside one decision. The coach needs a rotation option. The club needs to protect the transfer value of an asset on its balance sheet. The agent needs an uninterrupted season to negotiate a renewal. Those three needs push in the same direction, and convergent pressure always beats counter-pressure in a meeting room.

Then comes the part harder to repair than bone and cartilage. Fear of re-injury persists after physical testing has returned normal results. It changes how a player goes into a tackle, how he plants his standing foot, how he decides in a thousandth of a second out on the flank. Professional football has hundreds of metrics for speed and distance covered, but almost no metric for hesitation.

The result is that the second phase of a career — usually the more important phase — gets destroyed not by a bad tackle, but by a timeline chosen in a commercial meeting.

Gegenpressing has been decoded: when football becomes athletics

For years, high pressing was the biggest underexploited competitive advantage in modern football. Not any more.

One simple metric describes it: passes allowed per defensive action. The lower the figure, the more aggressively a team presses. When a handful of teams sit below eight, it signals an organised pressing system. When half a league sits below twelve, it signals that a trend has become the default.

The problem appears when a default meets a default. When two teams press, pressing stops being an advantage and becomes a cost. The ball goes longer, duels in midfield multiply, and the match settles into what I call a physical stalemate: twenty-two players running a great deal, the ball moving very little, and entertainment value falling in inverse proportion to distance covered.

In the mid-table group, I observe a different variant. These clubs lack the resources to build a three-layer pressing system, but they do have the resources to build a fitness machine. They turn football into athletics: run, duel, commit tactical fouls, break rhythm, wait for a set piece.

This works for one season. It also pushes injury load very high once a three-day match cycle becomes normal. A team running like an athletics squad in October is usually a crowded medical room in February. That is a testable prediction, and I still check it every season by comparing injury lists against the previous season's high-intensity running volume. Some seasons I am right, some seasons I am off, and I record both.

Tactics can be betrayed, but data cannot.

Esports and the speed at which betting money corrodes

There is one field I monitor with the same toolkit I use for football, and in it everything happens roughly three times faster: esports.

Here, the development structure is far thinner than in football. Youth academies in esports often amount to a room, a coach who doubles as a manager, and a short contract. At lower-tier events, player income sits well below the average cost of living while betting liquidity on those same matches is far from low. The mismatch between low income and high betting volume is the most familiar formula behind every match-fixing case in sporting history.

International integrity-monitoring bodies have for years published annual reports on the number of suspicious betting alerts, and a significant share of them come from esports disciplines. I am not putting a specific figure here, because those reports update quarterly, and a stale number in this field quickly becomes a wrong number.

My point is not that esports is more corrupt than football. My point is this: when a young sport's market scale balloons within a few years while its governance, player education and prevention infrastructure remain at an early stage, the rate at which competitive integrity erodes will be faster than in a sport with a hundred years of institutions.

Esports lacks academies, but it has an excess of the signals I learned to read from football.

The same set of signals: odds movement before kick-off, a young player suddenly underperforming in a match that does not matter, an unusual late substitution. I have seen every one of those patterns in football. Seeing them again in esports is an early warning, and an early warning nobody pays for is usually not acted upon.

The transfer window: where noise is paid for in signal density

Now back to this morning's file.

Nine of eleven fields are blank, but the context is not. We are in the middle of a transfer window. Every day brings hundreds of articles built to the same structure: an unnamed source, a big club, a player of unclear whereabouts, and a fee nobody confirms.

There is a real data structure underneath all that noise, and it sits in the places readers usually skip.

In Spain, contract release clauses are registered with the league and, mechanically, the player is the one who unilaterally triggers them. That structure explains why transfer rumours in Spain are often meaningless but the numbers are not. A release clause is a figure written into a document, not an opinion.

The second data structure is the wage bill. In a league with a revenue-based salary cap, no club can sign a player without first creating wage space. So the real order of operations in a major deal is always: sell first, buy second. News of a purchase almost always appears after somebody has sold. If you read that "club A is chasing player X" while club A has sold nobody and its wage bill is full, the probability of that deal is much lower than the headline suggests.

The third data structure is the amortisation of the transfer fee across the contract term. A large fee spread over five years does not hit the accounts the same way as the same fee spread over three. This is why contract structure — not just value — is often the single most important fact in a deal, and also the fact least often included in coverage, because it does not generate an attractive headline.

The thing I get from readers most during this period is not the question "who will win the league". They ask something else: is this story true. They need a reliability filter. And a filter only works if the person building it accepts saying "not enough data to conclude" about most of what passes through their hands.

Bias is the most expensive thing in the transfer market, and it has never appeared in a financial statement.

When a club buys a player on the strength of one performance in one big match, it is paying for the smallest possible sample. When it buys on the strength of a two-season development curve, it is paying for a larger sample. The market pays for small samples, then acts surprised at the failure rate.

The counter-intuitive angle: the industry rewards filling the blanks

This is the part I consider most important, and also the part that costs me the most readers.

If I filled the blank template this morning, I would have a nine-dimension piece, full of tables, three thousand words long, widely distributed. If I return a null result, I have a short document with no scoreboard, no characters, nothing to share. Content distribution systems do not evaluate those two products equally. Algorithms count engagement, not verifiability.

That is not a bug in the algorithm. It is a designed feature.

The deeper problem is this: the more detailed an analytical template is, the more dangerous it becomes when the input is empty. A template with nine dimensions and forty cells does not only contain questions; it also contains pre-made answers. The writer simply picks whichever pre-made answer fits the name that was inserted. That process does not produce knowledge. It produces a product shaped like knowledge.

In my trade there is one simple test I apply to every piece before filing: if every claim in this article is correct, what do I now know that I did not know before? If the answer is "nothing new", then the piece, however long and elegant, is decoration.

A transfer window in crisis will wipe out those who argue in bad faith.

I write that not to frighten anyone. I write it because I have watched it come true in both the literal and figurative senses. When the market contracts, clubs start re-reading their balance sheets, and everything that existed on tone alone disappears first.

Crisis does not create a new market; it strips the mask off the price-setters.

Systemic risk: three layers of contamination

Looking back at this morning's file with the eye of someone who has done data-checking work for years, I see three distinct layers of risk.

The first is analytical risk. If someone fills the blank template, they will generate club-level and player-level content out of nothing. That product will have names, numbers, and conclusions about injury risk, financial structure and league position. None of it will be verifiable, because there is no input to check it against.

The second is downstream risk. A document like this morning's is only safe if the input-status warning is preserved in every reuse. If that warning is stripped during editing, the remainder will be read as a completed analysis. In today's content pipeline, stripping that warning takes three seconds.

The third is recurrence risk. The "entities involved" field instructs the user to identify entities from the information points above, while the information point list is empty. That is a circular dependency. Every re-run of the template without fixing the extraction step will reproduce exactly the same failure, and with each repetition the chance that someone decides to fill the blanks increases.

Within those three layers, exactly one field in the whole file behaved correctly: the domain classifier, which returned "football". That small detail has diagnostic value. It shows the classification layer works and the fault lies in content extraction. When a system breaks, knowing precisely which layer broke is the entire difference between a one-time fix and a month of work.

I also have to say something about myself, because I know the weak points of people who have worked a long time. At 44, I have a tendency to trust the datasets I built with my own hands early in my career. That is a real bias. The only way I know to counter it is to periodically compare old predictions against actual outcomes, and when there is a divergence, to record the divergence rather than explain it away.

I also know another blind spot of mine. The role I assign myself as an archaeologist of academies makes me prone to seeing young players as a data structure, a development curve, a combination of metrics. That lens is useful, but it is missing something. After every quantitative table, I force myself to write a qualitative paragraph about the person: a 17-year-old living far from home, a family trading away a great deal, an injury that is not merely an entry on a list. Without that paragraph, my spreadsheet is only technically correct.

What to do next, rather than what to conclude

I will send this morning's file back to the desk with a short note: re-run the extraction step, and only reopen the analysis once the information point list contains at least three entries with named entities and specific timestamps. Attached to that is a second request: source metadata — outlet, journalist, publication date — must be a mandatory field. Without it, even a successful extraction leaves me unable to grade source credibility, and an analysis whose sources cannot be graded is not worth reading.

I propose one more metric for the process: the null-result rate. If an analytical pipeline never returns a null result, it is not analysing; it is manufacturing. A reasonable null-result rate is evidence that a system can refuse, and the capacity to refuse is the single most important skill of any evaluation system, from a newsroom to a scouting department.

In football we spend enormous resources measuring how far a player runs and what percentage of his passes find a teammate. We spend almost nothing measuring how large our own capacity to say "I don't know" is. That is the cheapest and most neglected metric in this industry.

This morning I could not write anything about a transfer. I could only write about why I could not write about it. To most readers, that is a less appealing product. To those eleven blank cells in that file, it is the only honest answer.

And in a transfer window, when thousands of articles a day assert things nobody can verify, an honest answer may be the scarcest thing on the market.