The Empty Data Frame: The Fabrication Trap in Sports Analytics

7:40 on a Monday morning

At 7:40 on a Monday morning, in the analytics office of an A-League club, a spreadsheet lands in the internal inbox. Subject line: Transfer Market Review, Week One. Fourteen columns. Club, position, minutes played, expected goals, transfer fee, wage-to-revenue ratio, contract expiry, agent, source reliability.

Thirteen columns empty. The fourteenth reads: A-League.

Thirty-eight minutes later, an intern returns the sheet with sixty tidy rows of data. The board signs off. Nobody asks where those sixty rows came from, because they look right. They have the right format, the right units, the right ordering, and most importantly they fill the gap before anyone can ask a question.

That was the moment I began to distrust the entire sports analytics industry. Numbers never lie, but the people who read the report do.

The story I want to tell today is not about a player, a club, or a match. It is about the frame. The empty frame, the beautiful frame, the frame with enough boxes to make people feel that everything is under control, while in truth there is nothing inside it at all.

The rumor economy runs on empty frames

Every transfer window, the sports industry produces a larger volume of information than in any other period of the year. Thousands of lines a week. Striker X has contacted club Y. Defender Z has refused an extension. Club W is preparing to sell a key player to balance the books. Half of it is true. A third of it is recycled old news. The rest is generated to fill space.

What catches my attention is not the volume of rumors but how they are packaged. Most transfer stories reach the reader as a pre-shaped frame: there is a selling club, a buying club, a fee, a contract length, an agent, a plan for how the player will be used. The frame is so complete that the reader has no room left to ask: which of these things has actually been verified?

In this profession there is a principle I learned rather late and paid rather dearly for. When a dataset looks too complete relative to the real resources of the person who built it, the most likely explanation is not that the person is brilliant. It is that the person filled the frame with guesswork.

There is a life-or-death difference between two kinds of input. The first is sparse information: there is an article, there are a few numbers, there is a statement, but the depth is missing. This can still be analyzed; you simply downgrade your confidence in the conclusion. The second is null input: no title, no source, no summary, no information points, no entities identified. This cannot be analyzed. And this is what very few people in the industry will admit, because admitting it means returning the report with two words: no data.

I do not believe in luck. I believe in numbers verified three times.

Anatomy of a null case

Imagine a two-stage data pipeline. Stage one does the deconstruction: it reads the article and extracts the title, source, article type, one-sentence summary, author stance, article purpose, list of information points, core viewpoints, entities mentioned, time sensitivity, source quality. Stage two takes that output and runs a nine-dimension deep analysis: technical, strategic, team and driver, competitive landscape, regulation and governance, transfer market, risk profile, public narrative, and industry transmission.

Now imagine stage one finishes and returns this: the domain label is filled in, correctly formatted, stating this is football. Everything else is empty. No title. No source. No summary. The information-point list is empty. The entity list contains a single line of instruction text: identify from the information points above. And above it, there are no information points at all.

This is a perfect null case. It is not a thin-data case. It is a case where the data never arrived.

What is frightening is its structure. The output still has every field. It still has a time-sensitivity field. It still has a source-quality field. They simply were never executed. The source-quality field contains an instruction rather than an assessment. The entity field contains a directive rather than a list. To a reader skimming past, those fields look as though they were processed and found not applicable.

The ambiguity between never executed and executed with no result is the most dangerous hole in the entire chain. It turns a technical failure into an analytical statement.

Based on my experience monitoring and handling club data, I would place the fault somewhere between the domain-classification step and the information-extraction step. The pipeline knew it was looking at football material, but it never read the body. Three plausible causes: the source sat behind a paywall, the link was dead, or the source was a non-text format such as video, a chart, or a results table.

In all three cases, the solution is not to guess. The solution is to refuse to publish.

When the domain label survives and the body vanishes

There is one small detail in this null case I want to linger on. The domain label was filled in.

In operational logic, a system classifying the domain correctly means it came into contact with content. You cannot tag a document as football if you have never seen it. But at the same time, the entire body appears in no other field. This suggests the content existed at some point in the pipeline, then disappeared between two steps.

For an operator, this is the hardest kind of fault to detect, because it makes no noise. The system does not throw an error. It returns a structurally valid result. If nobody checks the minimum character count of the body, the fault drifts through the entire chain and only surfaces when someone reads the final report and wonders why it feels so thin.

In football, the equivalent is a press conference where the coaching staff say they have analyzed the opponent in great detail, while in truth the opponent's match footage was never downloaded. Nobody lied outright. But the result of the match will speak for everything.

A low-tier contract can conceal a high-tier scandal.

Fabricating numbers by filling the frame

This is the part I want to give the most space to, because it is the real risk and it operates every single day of the transfer window.

When a language-generation system is handed an empty but fully structured frame, its path of least resistance is not to return a null result. Its path of least resistance is to fill that frame with the most plausible content it has ever seen during training. It will write a fluent football story with team names, player names, fees, and tactical analysis. And it will not flag that none of that content came from the source.

I call this phenomenon frame-filling. It differs from lying in that the writer has no intent to deceive. The writer is simply completing an assigned task. But the consequence is identical: a transparent gap becomes an undetectable falsehood.

In the sports industry, frame-filling happens at every level. A journalist short on sources will write that according to the information we have, the club is interested in the player, rather than saying plainly that no source has confirmed it. An analyst short on data will say the player has growth potential, rather than saying the sample is too small. A spreadsheet missing rows gets filled by linear interpolation, and nobody notes that three of the sixty rows are estimates.

What makes frame-filling more dangerous than lying is that it scales in batches. A club runs ten datasets a week. If one input source fails and the sheet is still filled, the same fault spreads to the other sheets in the same processing batch. By the end of the transfer window, the board is making decisions on a web of figures in which every number looks right but none can be traced to its origin.

Nothing on a balance sheet is ever a surprise.

Mariners, 68 percent, and the lesson of a short news item

I tell this story because it is the root of how I look at every dataset afterward.

In the summer of 2026, when I was a first-year broadcasting student at the University of Technology Sydney, I was interning in the sports desk of a Sydney radio station. One morning the editor asked me to write a short item: Central Coast Mariners selling striker Trent Buhagiar to Sydney FC for 250,000 Australian dollars. A three-hundred-word item. An intern's assignment.

I did not write three hundred words. I opened the Mariners' financial report and found a number that cost me weeks of sleep. The club was spending 68 percent of revenue on wages, while the A-League's recommended safety threshold was below 55 percent. That thirteen-percentage-point gap was not a minor detail. It was the real reason behind the sale of a player for 250,000 dollars.

I built a spreadsheet tracking the wage-to-revenue ratio of the entire league and wrote a two-thousand-word analysis. My lecturer received that instead of the prescribed short item.

The lesson was not in the 68 percent figure. It was that a three-hundred-word news item can contain a two-thousand-word financial story, and the writer of the short item only needs to not ask one more question to leave the frame empty. The empty frame here is the story of a player moving from club A to club B. The truth lies in the cash flow behind it.

Since then I have never written a surface-level item. Every piece I write begins with an anomalous number, then traces cause and consequence. And every time a dataset reaches my desk with every column filled, my first question is always this: which of these columns was filled with real figures, and which was filled with inference?

Mbappé and the iceberg we chose not to see

In the summer of 2026, while the whole world was talking about Kylian Mbappé after France's World Cup triumph, I spent an entire month building a young-player valuation model.

Mbappé was nineteen, scored four goals at the tournament, and his market value rose from 87 million euros before the tournament to more than 180 million euros after it. A rise of over 90 million euros in the space of a month.

What I wanted to test was not how good Mbappé was. I wanted to test how much of that value increase could be explained by actual on-pitch achievement. My model produced roughly 25 million euros of direct sporting value. The rest, more than 65 million euros, was expectation.

Mbappé is not the shock. He is the tip of an iceberg we chose not to see. That iceberg is the way the transfer market values human beings based on future narrative rather than past data. I compared Mbappé with Ousmane Dembélé and Marcus Rashford in the same window, on the same criteria of minutes played, goals, assists, and actual transfer fees, and concluded that the market was paying for expectation rather than achievement.

The piece was shared by a sports analytics account in Sydney, and a player agent contacted me asking for more of my data.

But the detail that matters more to me today is a different one. When I built that model, I had to fill many cells with estimates. Commercial value, media value, shirt-sales potential, brand impact. No source provided those numbers. I constructed them from my own reasoning.

If someone had asked me then what percentage of that model was real data and what percentage was frame-filling, I would have had to say I did not know. And that is precisely the problem. The value of a player lies not in his feet but in the way he is priced.

Morocco, 241 million euros, and the cost-efficiency coefficient

Four years later, at the 2026 World Cup in Qatar, I carried out an independent study of the spending efficiency of all thirty-two participating teams.

I took total squad value from transfer-market data and compared it with points won in the group stage, then extended it across each team's full run.

The most striking result belonged to Morocco. They reached the semi-finals with a squad worth around 241 million euros. For comparison, England's squad was worth around 1.87 billion euros, roughly eight times more in absolute terms. Morocco drew with Croatia, beat Belgium, eliminated Spain on penalties, and only stopped against France in the semi-finals. England went out in the quarter-finals.

I published a four-thousand-word report titled Cost-Efficiency Coefficients at the 2026 World Cup, arguing that tactical cohesion in defense generates sporting value the transfer market does not reflect. A football analysis site in Ukraine republished it.

But this time I did things differently. I wrote the two-hundred-word executive summary before writing the body. In that summary I stated three limits of the study explicitly: squad value is market value at the time of publication rather than the federation's actual cost, group-stage points are only a crude performance indicator, and a sample of thirty-two teams is too small to generalize into causal claims.

Those three limits were three cells I could have left blank to make the report look better. I chose to fill them with the truth.

The difference between a trustworthy model and a dangerous one is not the number of columns. It is whether the person who built it is willing to write down the things they do not know.

Wanderers, a five-month shutdown, and a twelve-month model

In 2026, the pandemic suspended the entire A-League for five months. I was twenty, finishing a graduation internship, when Western Sydney Wanderers contacted me to work remotely on a problem that had nothing to do with football: a liquidity crisis.

An empty stadium. Membership registrations down by 2,400 people. Matchday revenue gone. Fixed costs still there.

I built a twelve-month forecast model with three scenarios: optimistic, assuming football returned after two months; base case, four months; and pessimistic, assuming the whole season was cancelled.

The pessimistic scenario showed the club losing 7.5 million Australian dollars, far beyond the 5 million dollar provision the board had set aside. On the basis of that model, the board decided to negotiate a 25 percent wage reduction for key players.

There is one detail from that process I want to emphasize because it connects directly to the subject of this article. When I presented the model, one board member asked whether I could fill the missing cells with reasonable assumptions so the sheet would look more complete in the meeting with sponsors.

I refused. I left the cells empty and wrote beside each one: no data yet, will update when information arrives from the league organizer.

In a crisis, precision is the only thing that reassures all parties. A sheet with holes looks bad. A sheet filled with wrong numbers is worse, because it makes people decide on the basis of a reality that does not exist.

Football is emotion, but a club survives on algorithms.

Source tiering: the only thing that stops a rumor

Back to the current transfer window.

During a transfer cycle, the highest-value function of any analytical system is not predicting which player goes where. It is assigning a credibility tier to the source of a rumor. That sounds modest, but it is the only barrier against manipulation.

A rumor has value only when we know where it came from. An agent leaks to apply pressure on the current club in negotiations. A club leaks to inflate the price of a player it wants to sell. A social-media account leaks to drive engagement. A journalist with ties to an agent leaks to maintain that relationship. All four sources can produce an entirely identical piece of information, and all four can be wrong, or all four can be right for four different motives.

Without the source-tiering step, every rumor carries the same weight. And in a market where rumor is the main commodity, equal weight for everything means the biggest liar wins, simply because the biggest liar talks the most.

In the null case I analyzed at the start of this piece, the source-quality field contained an instruction rather than an assessment. That is what strips a system of its most important defensive capability. No source tier, no rumor weight, no way to distinguish a claim from an agent from a claim from a journalist who verified three independent sources.

For fans, this means they are consuming an information market with no price of value. For operators, it means a club can make buy-and-sell decisions based on the rumor that carries the highest weight only because it spread the widest.

When one failed source collapses a whole data batch

There is one aspect of frame-filling I want to give its own section, because it is rarely discussed in the industry.

When an input source fails and the result is still filled in, the fault does not stop at one report. It spreads across the entire processing batch of the same day, the same week, the same window. If a sheet has fourteen columns and thirteen are empty, the person building it has two options: return the sheet and say the data has not arrived, or fill it in themselves. The second option is always more attractive because it creates no friction.

But the consequence of the second option does not stop at that sheet. From the moment it is approved, every subsequent decision is built on that foundation. The transfer budget is allocated on that basis. The target list is drawn up on that basis. Negotiations with agents proceed on that basis.

At the end of the window, when the board sits down to review, they see a full web of figures. Every number looks reasonable. Every number has an origin that cannot be traced. And nobody can point to exactly where it went wrong, because wrong looks exactly like right.

This is why I always tell younger colleagues that the most important skill of an analyst is not building complex models. The most important skill is knowing when to return the sheet and say: I need more data.

The validation gate: refusing to publish a null result

There is a technical solution to this problem, and it is unbelievably cheap.

Before running any deep analysis, check whether the information-point list is empty. If it is empty, halt the entire process and return a null result. Do not allow any downstream step to auto-fill.

Check the minimum character count of the ingested body. A genuine football article cannot be shorter than a few hundred characters. If the body falls below that threshold, it is an ingestion error, not a concise article.

Add an extraction-status field with three values: success, failed, partial. This field must be mandatory. Then a field marked not applicable can no longer be confused with a field that was never executed.

Require every field to contain either a value or an explicit line stating that it was not assessed, with a reason. With that single rule, a whole class of misunderstandings across the chain disappears.

And finally, audit the entire batch before publishing anything from it. If one item in the batch is null, the sibling items carry a very high probability of the same defect.

None of these measures requires expensive technology. They require an organizational decision: to accept that a null result is a valid result. That is the hardest part, because in sports, a null result is treated as failure.

The counter-intuitive angle: more data, more fabrication

This is where I want to turn suspicion on my own industry.

Over the past decade, sports has invested heavily in data. Sensors in shirts, positional tracking systems, machine-learning injury-prediction models, expected-goals metrics, real-time player-value indices. In theory, more data should reduce the amount of fabrication.

In practice it goes the other way. More data means more room to fill frames. An analyst holding twenty metrics will not say I do not know. That person will pick the three metrics that support the desired conclusion, ignore the other seventeen, and present a conclusion that looks grounded.

Much the same happens at the media layer. The pressure to produce new content every day, every hour, every minute during a transfer window gives writers an incentive to fill the frame rather than leave it empty. A piece saying there is no new information will not be shared. A piece saying the club is considering three options will be shared thousands of times, even if those three options were generated in ten minutes.

This is the point I want you to sit with. The problem is not whether data is scarce or abundant. The problem is that this industry has never taught its practitioners how to say the words I do not know.

And there is a subtler risk. When an analytical system is designed to always find a new angle, a new discovery, a piece of information nobody else has, the pressure to produce a discovery outweighs the pressure to respect the truth. A pipeline with no capacity to return a null result is a pipeline designed to fabricate.

I recognized this in my own work. Last year, when I was preparing a five-year impact assessment for an expanded global club competition, I spent six weeks refining the model. The model showed a potential profit of 12.8 million Australian dollars if the club invested 3 million dollars a year in its youth academy and sold players to Europe.

I kept revising the assumptions to achieve greater precision. The report was three weeks late. The board was displeased, though it acknowledged the content had value.

What I learned was not to abandon perfectionism. What I learned was not to use perfectionism as an excuse to delay admitting the things I did not know. A model that is eighty percent right and delivered on time is worth more than a model that is one hundred percent right and never reaches the person who needs it. But a model filled with assumptions to look one hundred percent right is the most dangerous of the three.

There is another temptation I want to name: the temptation to reverse the conclusion at any cost. When the majority says a player is valuable, analysts tend to want to prove the opposite to demonstrate independence. I have been guilty of this. But if the data stands with the majority, then trying to prove the majority wrong is not independence. It is another form of frame-filling, differing only in that it fills the frame with contrarianism instead of consensus.

What I think will shape the next transfer window

I do not think the next competitive advantage in sports lies in who collects more data. Data has become a commodity. Everyone has access to the same set of figures on minutes played, goals, assists, and market value.

The next advantage lies in discipline. The discipline to refuse to publish when the data has not arrived. The discipline to mark clearly which cell is a real number and which is an estimate. The discipline to assign a source tier to every rumor before sharing it. The discipline to say the words I do not know in a room full of people waiting for an answer.

In this transfer window, when you read a story about a transfer, I suggest you ask yourself three questions. First, where does the transfer fee figure come from, and what motive does the person who gave it have. Second, if you remove the fee, how much genuinely verified information remains. Third, and most importantly, if the writer had no information at all, would that article still exist.

If the answer to the third question is no, then that article is not information. It is an empty frame filled in before you had a chance to see it was empty.

An honest empty frame is worth more than a full dishonest sheet. And in a market where thousands of lines of data arrive every week, the person who can stand still and say I do not know this part is the person who controls the game.