Nine Analytical Dimensions, Zero Data Points: Notes From a Numbers Reader

At 3:12 in the morning I reopened the file my partner had sent. Nine analytical dimensions. Each with an assessment table, a risk-flag column, numbered conclusions. It was clean enough that printed out it would look like a strategy brief worth paying for.

The actual data inside it: none.

The title field read N/A. The source field read N/A. The list of information points was an empty array. The entities field contained a raw instruction — "identify from the information points above" — while the very list it referred to was blank. Of eleven input fields, exactly one carried content: the domain label, reading "esports."

Eleven fields. One correct. Nine analytical dimensions still came into the world.

Seventeen years of reading tables taught me this is the hardest kind of failure to catch. Wrong data can be argued with, because there is something to argue about. Data that never existed, run through a system that keeps going, cannot be argued with at all.

I work as a transfer market administrator, based in Hanoi, writing about esports for Vietnamese readers. The daily job is taking raw material — match records, transfer valuations, fitness tracking data — and turning it into something a decision can rest on. My process has two steps. Step one extracts events: who, when, where, how much, under what conditions. Step two places those events inside a model: which patch, which format, which roster, which cash flow.

Step two lives entirely off step one. Without step one, step two is decoration. And decoration, when formatted well enough, is more deceptive than a wrong number.

The report from the night before was exactly that. It invented no team, no player, no tournament. It wrote N/A in every substantive position and stated plainly that nothing could be assessed. Technically the model behaved correctly: it refused to conclude without premises. Yet it still produced nine sections, still numbered its conclusions, still set every line in type with rules and boxes. A hurried reader sees "nine analytical dimensions" and assumes nine dimensions of work were done.

The most telling part was the entities field. It held a command instead of data: identify the entities from the information points above. That list was empty. This is a self-referential loop — the system instructing itself to find what it was never given. The inevitable result is nothing. But because the field contained words, it never got flagged as blank in any summary.

I have watched enough matches to know what good data looks like. In 2026 I built an xG model from 26 rounds of V-League and found Long An averaging 0.72 expected goals per match, lowest in the league. The editors said football is not mathematics and refused to publish. Long An were relegated at the end of the season. I was rejected in 2026 over a model. Seven years later I get paid to write about it.

A year later I calculated PPDA across all 32 teams at the 2026 World Cup and found Croatia at 9.8 — very low, meaning they did not press continuously. Yet their success rate per opponent pass pressed led the tournament at 23 percent. The article predicting Croatia in the final was mocked, then vindicated. In 2026 I counted Morocco allowing opponents an average of 4.2 touches in their own box per match through a disciplined low 5-4-1; against Portugal, Sofyan Amrabat recorded six successful tackles and nine ball recoveries. Every one of those pieces stood on the same foundation: somebody sat down and counted.

Those three stories differ in tournament, timing and conclusion. They share exactly one thing: real data, and a person accountable for the counting. The nine-section report had neither. It was the shape of an analysis without the body of one.

An empty model still produces conclusions that are formally correct and substantively false; that failure is harder to detect than bad data, because there is nothing to check it against.

One match is a story. Fifty matches are the truth. Here the sample was zero, and what got delivered was a book with a complete table of contents and no printed pages.

The counterintuitive part is that everyone blames the model. The reflex is to say the system hallucinated, that it made things up. Look closely and it made nothing up. It wrote N/A everywhere it should, self-flagged low confidence, and even left a warning that N/A here means unknown and must never be read as no problem found. The model did its job. What failed was the gate in front of it: no null-check halted the chain when the input came back empty. An automated pipeline without a safety valve will mass-produce fluent, confident, baseless analysis at industrial speed.

The second counterintuitive angle cuts closer to my own trade. In Vietnamese sports media we rarely write N/A. We write "reportedly," "a source close to the situation," "highly likely," "according to expert assessment." Those phrases do one thing: they fill the hole with prose. They are N/A in a suit. And because they are well dressed, nobody ever checks them.

When I delivered that pay-cut advisory, they looked at me like a man without feeling. I was handing over data, not emotion. Now I read it the other way round: the genuinely cold act is delivering a report with no data inside it, beautifully bound, letting the recipient believe they were advised. Handing over data — even empty data — respects the reader. Handing over the appearance of data does not.

One detail I keep as a professional note. This failure has a reproducible fingerprint: the domain label populated, while title, source, article type and information points are all blank, and the entities field still holds template instruction text. Four signals appearing together is not chance. It points to a process that stopped right after domain classification and before content extraction — meaning the source article is almost certainly intact and can be re-extracted. In this trade, knowing where a fault sits matters more than knowing how severe it is.

For a national team or a club, the same logic holds. A scouting report that calls a player "full of potential" without minutes played, distance covered per match, or injury frequency is not a scouting report. It is an advertisement line. Even a nine-figure contract begins with a small note about minutes played.

The signal I will track next cycle is not in anyone's conclusions. It is in the input table. Extraction success rate across all submitted articles. The share of entity fields filled with real data rather than template text. The null rate on titles and sources. Whether the source-quality and time-sensitivity fields are actually assessed. Those four numbers cost far less than reading nine dimensions, and they tell me in advance which dimensions are worth reading at all.

People ask where I put my trust in an analysis. I trust the input, because it is the only part that cannot be decorated. Conclusions can be good or bad. Inputs have two states: present or absent.