Empty Data, Full Conclusions: Notes from a Transfer Analysis Room in Busan

The office in Busan, 2:40 in the morning. I open a fourteen-page scouting file a colleague sent over. The column headed "verified data" is empty: no minutes played, no chance-creation rate, no positional map, not even a verified date of birth. The column headed "conclusion" is full. Seven hundred words asserting that this player suits a high-pressing system, reads the game well, has leadership qualities. Not one line carries a source.

I read it three times, then did the thing I always do when I am suspicious: I counted. Fourteen pages, twenty-two assertions, not a single independently checkable fact. That file was not an exception. It is the pattern — and that pattern is pricing players, shaping squads and deciding who plays next season.

Misjudging a player costs you one place in the lineup. Misjudging how you know a player costs you the entire process.

That night I did not edit the file. I wrote a different page listing exactly what it would need to become a real file: four hundred minutes of footage from the domestic league, positional data from the last three matches, contract status confirmed by the club. That page was longer, drier, and almost nobody wants to read it.

Context: extract first, interpret second

Professional sports analysis runs on two clearly separated steps. Step one is fact extraction: which team, which player, which league, which time frame, which numbers, which source. Step two is evaluation, comparison, forecasting. Every credible report passes step one before it is allowed to enter step two.

When step one returns nothing — no team, no player, no competition, no date — step two has to stop. Not because the writer is incompetent, but because every assertion written afterwards has nothing behind it. The correct output in that situation is an analytical framework with cells marked "insufficient information", plus a concrete list of what is missing.

Applied to football, that framework has nine layers. The layer of seasonal tactical change — the equivalent of a patch in esports — determines which metrics are still readable and which have expired. The competition-format layer determines how noisy results are: a double round-robin league produces a far cleaner signal than a single-elimination cup. The squad and player layer. The regional layer, meaning the balance of power between confederations. The club finance layer. The rules and governance layer. The risk layer. The media-narrative layer. And the layer of transmission across the wider industry.

Each layer has one root question: which facts can be verified, and who verified them? Skip that question and what remains is prose. And prose does not hold a starting place for a player, and does not save a club from relegation.

The core: four times I learned what data costs

In 2026 I was a first-year student in Busan, collecting K League 2 match data by hand. Asan Mugunghwa were top of the table. I calculated their xG: 1.02 expected goals per match. Busan IPark, below them, were at 1.48. That gap was not spread thinly across many small metrics — it was concentrated in a single point: Asan were awarded six penalties in six consecutive matches. I wrote a post on my personal blog predicting they would fall away in the second half of the season. It reached 2,000 views, the biggest number my student blog had ever seen. Asan finished fourth and lost in the play-offs.

Do not trust the table, ask xG. The table tells the past, data tells the future. There is one more line I repeat to colleagues whenever I read a summary sheet: a team scoring penalties in 6 of 6 matches is not playing football, it is playing luck.

In June 2026, in Kazan, I analysed South Korea's 2-0 win over Germany. Germany's PPDA was 5.8 — an extremely aggressive press on the conventional scale. Many commentators used that number to criticise Shin Tae-yong's approach. I split the data into fifteen-minute windows and found Germany ran their highest volume between minutes 60 and 75, then their pressing structure broke apart after Kim Young-gwon came on. South Korea needed only three shots on target to score twice. I wrote a rebuttal and took heavy criticism in the comments, with some arguing I was defending a lucky win. Three weeks later FIFA published its technical report confirming exactly what I had measured. I drew a professional rule from it: I was once attacked for daring to question PPDA. FIFA confirmed it. And a second line, written for myself: PPDA of 5.8 sounds frightening, but a team out of gas at minute 75 is the genuinely frightening thing.

In the summer of 2026, national leagues had to play in empty stadiums. I tracked 214 matches in the Bundesliga and K League 1 from May to August, logging each match on the same form, with the same variable definitions, to avoid quietly changing my own yardstick mid-study. The home-win rate in the Bundesliga fell from 43.2 percent to 37.8 percent; average goals rose from 2.79 to 3.12. I published the results on Medium, and an editor at Football Analysis reached out, specifically about mining GPS positional data from Korean clubs — paid data I had never had access to before. It was the first time I wrote for a publication with an editor, which meant the first time I had to standardise presentation: comparison tables, source footnotes, neutral language, and no more self-appointed blogger voice. I started from a student blog with 2,000 views. Data does not care who you are, only whether you read it correctly.

In June 2026 I proposed signing Lee Kang-in from Mallorca for 8 million euros. My data showed him in the top ten in La Liga for chance-creating passes per 90 minutes, at 2.8 — higher than Isco. The board rejected it, citing insufficient defensive output. I registered my dissent and accepted the decision. Six months later Lee Kang-in shone and helped Mallorca stay up, while my club finished eighth. I collected every email, data report and meeting minute and wrote a fifteen-page internal analysis for the board, identifying a process failure rather than blaming any individual.

Those four stories share one thing. In all four, I had to go and gather raw data myself before I allowed myself to write the first conclusion.

A professional footnote on esports, where the silence of data is more dangerous

I work as a transfer market administrator specialising in esports, and this is where the lesson above doubles in price. In football, a metric such as chance-creating passes per 90 minutes holds relative meaning across seasons, so a 2026 player can be compared with a 2026 player. In esports, a single patch can invalidate a metric within two weeks. Pick rate, ban rate, win rate, average kills — all of them are bound tightly to the live game version. Carrying figures from the previous patch into the next without a version note is systematic bias, not random noise you can shrug off.

The less data there is, the easier it is to fill the gap with adjectives. When a player's sample is three matches, people write "in great form". When it is three hundred matches, they are forced to write "participated in 68 percent of fights in the first twelve minutes". Three matches and three hundred matches are not the same kind of evidence, and an honest report must state where on that scale it stands.

This is why I never import a football metric into esports without localising it first: explaining what the original variable measures, how it behaves in the new environment, and the limits of the conversion. Dropping xG straight into a five-versus-five title is a beautiful metaphor on a slide and a mistake in a report.

The counter-intuitive angle

The market pays for certainty, not for accuracy. A report saying "this player will succeed" is read ten times more than one saying "insufficient information to conclude, we need four hundred more minutes". Writers get pushed toward confidence, because confidence gets traffic and caution does not. Traffic then decides who gets invited to write again.

But here is the paradox I have encountered often enough to believe: an empty result is the most useful result. When the fact-extraction step returns zero, the information gained is not "there is nothing here" but "our data pipeline is broken" or "our monitoring network has a blind spot precisely in this market". Silence from data is still a signal — just a far harder one to sell than an attractive prediction.

There is a trap attached that I once fell into: missing data does not mean missing risk. A blank line under wages can mean the club pays on time, or it can mean nobody bothered to check. Both situations produce the same white page and two opposite conclusions. A clear-headed analyst must write both possibilities instead of choosing the comfortable one.

There is another trap, and it is about temperament. After the PPDA episode in 2026, I realised I had a tendency to become defensive when criticised, and that tendency ruins rebuttal. Drawing a sharp line between a personal attack and a methodological challenge is a survival skill. Anyone calling me arrogant does not need an answer. Anyone offering a different time window that flips the conclusion needs an answer the same day.

What I am taking into next season

I am proposing a small change to our internal process: every report must carry a data-density index — the ratio of independently verified facts to assertions. A fourteen-page file with twenty-two assertions and no facts has an index of zero, and an index of zero means no conclusions are permitted, whoever the author is, whatever their title, however many matches they claim to have watched.

The coming transfer window will again be full of confident analyses of players nobody has watched for a full ninety minutes, and of league tables cited as though they were causes rather than outcomes. Before believing any line in them, I will ask one very simple question: which number stands behind this line, and who measured it.