A Nine-Page Report With No Subject: When Sports Data Analysts Fabricate Their Own Reality

Last November, in a data team meeting in Seoul, a young colleague pushed a nine-page esports analysis report across the table to me. 'Read it,' he said. 'It took me three days.' I turned each page. Patch Impact Assessment. Roster Assessment. Regional Strength Comparison. Risk Matrix with seven risk categories. Clean layout, balanced tables, measured language.

On the third line of the first page, I stopped. 'Game Title: N/A - insufficient information.' I kept reading. 'Version/Patch: N/A.' 'Tournament Name: N/A.' 'Analysis Subject: N/A.' 'Regional Tier: N/A.' Nine pages, and not a single game, team, player, or tournament was named. It was not an analysis. It was an empty skeleton, filled with N/A and methodological notes.

My colleague didn't understand why I put it down. 'I followed the process,' he said. 'I didn't fabricate numbers, I clearly wrote there was no information.' That was precisely the problem. He thought being honest with data meant writing N/A in every empty cell. But in our trade - the trade of telling stories through numbers - there is a trap more dangerous than fabricating numbers: fabricating the subject. There are matches the naked eye cannot see; the spreadsheet must tell them. But when the spreadsheet has nothing to tell, silence is also an answer.

In professional sports data analysis, we operate on a two-stage model. Stage-1 extracts: reading the source article, pulling out information points, identifying entities (teams, players, tournaments, patches), summarising author stance, assessing time sensitivity. Stage-2 takes that output and interprets it with domain expertise: patch analysis, roster analysis, club finance, risk, industry transmission.

When Stage-1 returns an empty result, the Stage-2 analyst faces a fork. One path is to stop and report there is nothing to analyse. The other is to infer a subject - guess which game, team, or tournament the article might have been about - and then write an analysis that looks complete. The second path leads to disaster. And that nine-page report proved it is not merely a theoretical risk.

Vietnam's esports scene faces a smaller-scale version of the same problem. Watching VCS, the Vietnam Championship Series, across seasons, I noticed many 'analyses' widely shared that were really just match recaps in florid language. They look like analysis because they contain judgments, predictions, advice. But they contain no data. More importantly, they contain no measurable subject - nothing that could be proven wrong.

To understand why a fabricated subject is more dangerous than a fabricated number, you need to understand the architecture of a proper analysis pipeline. Here I have to thank the two-stage model I was trained on: it forces me to confront emptiness before I am permitted to interpret. A spreadsheet does not lie; readers must learn to listen. A wrong number can be caught by another number. But a wrong subject - a team that does not exist, a patch never released, a tournament imagined into being - has nothing to be checked against, because the entire frame of reference was warped from the foundation.

Subject substitution is the deadliest failure in an analysis pipeline. When Stage-1 is empty, analysts tend to fill the void with inference: they read the task title, glance at surrounding context, tell themselves 'this is probably about League of Legends' or 'this is probably a VCS team', and then write an analysis that reads convincingly but is built on sand.

I have seen this in subtler forms. In 2026, analysing Germany's 0-1 loss to Mexico at the Russia World Cup, a male reader commented: 'Girls should not speak about tactics.' Instead of arguing, I published a new post with an xG chart: Mexico generated 1.8 xG, Germany 0.9. Numbers do not defend; they simply speak. Without them, I would have had to infer - 'Mexico was lucky' or 'Germany was complacent' - and every conclusion after that would have been fabrication. Don't argue with words; let xG speak.

My 2026 case was even cleaner. At 14, I sat by the pitch at the Seoul Youth League with a notebook, recording data from FC Seoul U-18 versus Anyang U-18. Midfielder Park Ji-ho had 92% pass accuracy - a gorgeous number. But when I counted forward passes, the figure was three. I wrote in my report that his midfield control was 'soulless' because it lacked line-breaking passes. FC Seoul's coach confirmed the assessment and used it to adjust tactics. That was the first time I saw data tell a truth the naked eye missed.

But what if I had not had that notebook? If I had only heard that Park Ji-ho passed well, and then written an analysis about 'the importance of the controlling midfielder' - I would have created a truth that did not exist. The subject would not have been fabricated, but the subject's attribute would have been. That is subject substitution at the micro level.

At the macro level, the consequences multiply. Imagine an analyst receiving an empty Stage-1, and under deadline pressure 'assuming' the article was about a MOBA title. They write a Patch Impact Assessment for a patch that does not exist. They build a Regional Strength Comparison for two unidentified regions. They lay out a Risk Matrix for seven risk categories with no evidence that any risk is present. The report is formally complete but contains not one verifiable sentence. And if anyone believes it - an investor, a coach, another journalist - they are making decisions based on fiction.

Screening asymmetry is why the risk layer of this industry is especially dangerous. The most severe risks - unpaid wages, match-fixing, key-player injuries, publisher sanctions - are silent by default. They surface only when actively screened for. When Stage-1 is empty, it means the screens were never run. The absence of a risk signal is not evidence of safety - it is evidence of non-inspection.

I learned this lesson as a contract data reporter. In 2026, before South Korea faced Portugal at the Qatar World Cup, I analysed South Korea's PPDA across four group-stage matches and found it rising from 10.5 to 7.8 in the first 30 minutes of each game - meaning they raised their pressing line dramatically after kickoff. Before the match, I predicted South Korea would press hard from the start. In reality, they recovered the ball 11 times in Portugal's half in the first 30 minutes, and the decisive goal came from a pressing situation. When I predict, I do not look at emotion; I look at PPDA.

But my point is not that I was right. My point is that without PPDA data, I could still have written a very plausible prediction - 'South Korea will fight for honour', 'Portugal might be complacent' - and if they won, I would be praised for insight. If they lost, I would say 'unlucky'. That kind of analysis has no value because it cannot be wrong. An analysis that cannot be wrong is not analysis. It is divination.

Here I touch a large paradox of the sports data industry. We live in an era of exponentially growing data - every match generates millions of data points, every player tracked down to each movement. But more data does not mean more information. It only means more opportunities to produce analyses that look objective but are really guesses wrapped in the language of numbers.

Framework completeness illusion is the trap the nine-dimension framework itself creates. A report containing Patch Analysis, Tournament Analysis, Team Analysis, Regional Analysis, Finance Analysis, Governance Analysis, Risk Profile, Narrative Analysis and Industry Transmission will convince a non-specialist that it has covered every angle. But if each dimension contains only N/A, formal completeness is disguising substantive emptiness.

I shared this phrase at a small Seoul workshop early this year, and a senior editor pushed back: 'If data is missing, writing N/A is honest, better than writing nonsense.' I agreed halfway. Writing N/A is indeed more honest than writing nonsense. But the problem is this: a nine-dimension framework full of N/A can still be circulated, cited, and used as a finished analysis product. Readers skimming headlines and section titles will not read every N/A line. They will assume that a report this long and structured must carry proportionate information. That is why I set a rule: if the subject has not been established, the correct product is not a nine-dimension report but a short 'out-of-scope' notice.

Back to my young colleague. I asked him: 'If you strip out all the N/A sections, what is left?' He was silent. I continued: 'Right. Nothing. But that is not your failure. That is the pipeline's failure. Stage-1 broke, and you were handed an empty box. The right move is not to decorate that box, but to go back and check why it was empty.'

We checked. The source article had failed to load - the server returned a 403 due to a paywall, and the extraction script silently fell into a safe mode that returned an empty structure instead of an error. If we had not paused to question the report's integrity, the team would have written hundreds of lines of analysis about a subject that did not exist.

At industry scale, this is not an isolated story. From my seven years in sports analysis, most errors do not come from misreading data. They come from failing to check whether the data exists before interpreting it. Data gaps are invisible until exposed. The only way to expose them is to build a process that forces emptiness to become visible, instead of being swallowed into a confident-sounding answer.

I do not believe in luck. I believe in blocked shots and forgotten gaps. In this case, the forgotten gap was the subject of the analysis itself.

There is a detail I have not told. That nine-page report had a section titled 'Highlights & Opportunity Identification'. In it, the young analyst wrote: 'A clean diagnostic opportunity: the failure is total rather than partial, making it easier to diagnose than a degraded-extraction case - where some fields are right and some wrong, and errors hide in the correct-looking ones.' I read that line and had to admit: he understood the nature of the problem better than I had thought.

The scariest thing about partially corrupted data is not the corrupted part. It is the part that looks intact. In our trade, a correct number next to a wrong number is more dangerous than two wrong numbers next to each other, because the correct number creates trust, and that trust transfers to the wrong number. That is why when I audit a K League dataset, I do not just count empty cells. I count cells with data. I cross-check at least two independent sources. I record sample size, confidence level, and assumptions. And I always leave one cell for the question: 'If this number is wrong, how would I know?'

In fact, most of the value in my work is not in drawing conclusions. It is in knowing when not to. In 2026, when global football halted due to the pandemic, I was 17 and spent the time collecting K League 1 data from 2026-2026, computing PPDA for every team. The result showed Ulsan Hyundai pressing very effectively with a PPDA of 8.2 - meaning they allowed opponents at least 8 passes before recovering the ball. I predicted Ulsan would dominate the coming period. When football returned, they went unbeaten in five matches. The piece was republished by Sports Donga.

But what I am proud of is not the correct prediction. It is that I clearly noted in the article: a two-season sample, differences in fixture schedules, and the possibility that mid-season coaching changes distorted some metrics. I did not present the result as truth. I presented it as a conditional model.

That spirit - the spirit of the conditional model - is exactly what the nine-page report lost. It was humble enough to write N/A in every cell. But it lacked the courage to conclude: when there is no subject, the entire analytical framework must be demoted to a short diagnostic note, not a finished product.

They told girls not to talk tactics; I drew charts instead of answering. In this case, the chart I wanted to draw was an empty one - a two-axis graph with no data points. And I wanted to place it on the first page of the report, so that anyone opening it would immediately see: what should be here is not here.

The sad truth is that the sports analysis industry rewards form. A long report is rated higher than a short one, regardless of content. An article with many charts is shared more than one with a single correct chart. In that environment, the nine-page report full of N/A is a rational product of a system that incentivises the wrong thing. The writer is not at fault. The process designer is - and so are we, as long as we keep consuming products that look professional without asking about the subject.

My professional view on data: the heatmap has become the new divination. It conceals a player's real role in a tactical system. And at a deeper level, the heatmap also conceals the emptiness of the data - a beautiful heatmap can be generated from entirely unfounded data, as long as the presenter knows how to blend colours. The same is true of Patch Impact Assessment. The same is true of Risk Matrix. Any analytical structure that can be filled with language without data can be filled with fiction.

So I set three questions for any analysis that passes through my hands. First: who is the subject, and is it measurable? Second: if the main conclusion is wrong, what evidence would tell me? Third: if I strip away the decorative language, is what remains enough for someone else to reproduce the conclusion? The nine-page report failed all three. It had no subject, it could not be wrong, and it could not be reproduced - because beneath the language, there was nothing.

In football, people often say the best defence is one that never faces a dangerous attack. In data analysis, the best report is one that knows it should not exist. A mature analyst is not only someone who can read numbers, but someone who knows when the numbers are not yet enough to speak. And that skill - the skill of falling silent at the right moment - is far rarer than the skill of presenting a beautiful table.

Looking toward the next analysis cycle, I believe the biggest challenge will not be collecting more data. It will be distinguishing real data from data generated to fill gaps. When language models can produce a nine-page report in seconds, the analyst's value will no longer be measured by article length, but by the ability to detect that the article had no subject from its very first line. Anyone can produce words. Not everyone can produce truth. And a good analyst, standing before an empty skeleton, does not rush to stuff flesh into it, but stops and asks: whose skeleton is this?