Nine Analytical Dimensions, Eight Left Blank: When an Analyst Refuses to Invent Numbers

Nine Analytical Dimensions, Eight Left Blank: When an Analyst Refuses to Invent Numbers

At 1:12 a.m. I opened a table tennis analysis file sent by a data group. The file contained exactly nine sections, numbered one through nine, following a rigorous framework: technique, tactics and equipment; player data and head-to-head records; tournament systems and scoring rules; the international competitive landscape; rules and governance; coaching staff and talent pipelines; the risk surface; public narrative and expectation; and finally, industry transmission.

All nine sections carried the same line: insufficient information, cannot assess.

There was not a single player name in the file. Not one score. Not one metric, not even the most elementary one, such as points won on serve. The only field fully populated was the domain label: table tennis.

Ten years ago I would have opened the conclusion first. That night I opened the input audit first, and it turned out to be the only part of the report worth reading.

The production pipeline of an analysis

A piece of sports analysis passes through four stages: source collection, information extraction, analysis, publication. In football nations with dense data infrastructure, each stage has a separate owner. In Vietnam, the first three usually belong to one person, and that person is usually squeezed by the deadline of the fourth.

I know that pressure from the other side of the trade. Eleven years managing the transfer market taught me that a player file missing physical data still has to go before the committee on time. Nobody lets you submit a blank page. So people fill the gaps with instinct, with a rumour from an agent, with a memory of a match three years old. And once the file is approved, nobody traces back what the gaps were filled with.

Vietnamese football has paid tuition for that habit. In 2026 I bet on xG. The V-League answered with a shock. In the match where Hanoi FC beat Thanh Hoa 3-2 at Hang Day, InStat data recorded Hanoi generating 0.9 xG while Thanh Hoa generated 1.7. The next day the press crowned the home coach a tactical genius. I wrote that such a conversion rate could not hold. Four rounds later Hanoi dropped points repeatedly and nobody mentioned the praise again.

In the summer of 2026 a major football site asked me to predict the World Cup champion with my own model. I put Brazil on top based on group-stage total xG and PPDA. Belgium eliminated Brazil in the quarter-finals; France won. Reviewing match by match, I found the error lay somewhere quite different from where I had assumed: France cut their PPDA from 11.2 in the group stage to 8.7 in the knockouts. The champion changed how it played at each stage, while I applied one fixed number to every stage. World Cup 2026 taught me: data is never a single layer.

Those two episodes left a professional habit. Before analysing a match, I check whether I actually have the raw material to analyse.

The report that night was the first time I saw someone do that openly.

Auditing the input layer: the work nobody wants

The report devoted its entire opening to a twelve-row audit table. It listed what should have been there: the original article title, the source, the article type, the list of information points, a one-sentence summary of the core argument, the author's stance, the purpose of the article, the entities involved, time sensitivity, source quality.

The status column held only two values: missing, or unresolvable.

Information points was the most important column and also the emptiest. Under this framework, every downstream conclusion must trace back to a specific information point, by number or by content. With zero information points, the citation chain breaks at the very first link. The analyst wrote it plainly: with an empty list, reconstructing that chain is structurally impossible.

The audit also flagged a small but telling administrative error. The entities field carried an instruction: identify from the information points above. That pronoun has no antecedent. It points into empty space.

And this is where I slowed down. The report did not fill the gap. It stated outright that producing technique breakdowns, head-to-head tables or risk matrices here would mean inventing players, inventing records, inventing match contexts. It added a line I wanted to frame: a confidence label on an invented fact is worse than no analysis at all.

In the match-report trade, that line is a declaration of war. Because the industry default is: there must be a piece. The piece must have numbers. The numbers must look specific.

Why the industry defaults to invented numbers

I do not believe Vietnamese sports writers enjoy fabricating. I believe they are placed inside a reward system that pushes them there.

First is frequency pressure. A V-League round has seven matches in two and a half days. Each match needs at least one piece. Each piece needs a new angle. The number of genuinely new angles the data permits is smaller than the number of pieces required. The gap gets filled with adjectives.

Second is the search optimisation standard. Recent algorithms reward content with information gain, meaning it must bring something the reader did not know. Sounds right, but it creates a paradox: when you have no new data, the cheapest way to manufacture information gain is to manufacture a new interpretation of old data. Interpretations cannot be verified. Numbers can. The system therefore rewards the unverifiable.

Third is the habit of copying foreign models wholesale. Raw metrics like xG, PPDA and progressive passes arrive in Vietnam without a user manual. A reporter who reads the xG table of the Premier League and applies it directly to the V-League will see a tournament that looks very different, and instead of doubting the model, he doubts the tournament.

Fourth, and hardest to admit, is the writer's instinct to avoid gaps. A twelve-paragraph piece needs all twelve paragraphs to flow. One gap makes the writer feel under-skilled, even when the gap belongs to the data and not to him.

The transfer market administrator does not manage cash flow. They manage expectations. I think that sentence transfers intact to this trade: the sports writer does not manage truth. They manage the newsroom's expectation that the piece will be full.

The broken citation chain

On the surface, a V-League analysis looks dense with numbers. Trace each number back to its origin and the picture changes.

The most familiar phrase is according to statistics. No provider named, no date, no sample scope. A sentence like the home side had 62 percent possession could come from three different sources producing three different results, because the definition of possession depends on whether dead-ball time is counted.

Layer two is the heat map. I have said many times that the heat map has become a new form of fortune telling. It feels strongly scientific, but a red streak down the left flank does not tell you whether the player chose that position because of a tactical instruction or because the opponent forced him there. A heat map records where a player's feet landed, not why.

Layer three is imported advanced metrics. We once used a European provider's metric set to evaluate a midfielder in the V-League. The metric scored him highly because it rewards passes played forward toward goal. But his team played counter-attacking football and he only passed once he had three square metres of space. The metric was not wrong. The metric simply did not know what system he was playing in.

Those three layers combine into a citation style with no accountable owner. When the number is wrong, nobody is held responsible, because nobody claims to have produced it. The chain breaks precisely where in theory it should be strongest.

Three failure causes, and their V-League map

The report offered three hypotheses for the empty input layer, each with a confidence label. I find all three describe fairly accurately what happens weekly in my trade.

Hypothesis one, highest probability: the upstream extractor returned null. It received a text it could not classify and emitted the template skeleton without filling any field. In Vietnamese football, this is the press release case. A club sends three paragraphs with no numbers, only adjectives. The content machine still has to run. The result is an analysis with nothing to analyse.

Hypothesis two: the pipeline dropped the body. The source fetched empty, or sat behind a paywall, or was taken down before the extractor could read it. For Vietnamese table tennis this is routine. National championship results are often published as a scoreboard only, with no game-by-game record, no serve statistics. The writer is forced to reconstruct the match from memory.

Hypothesis three: schema mismatch. The information points exist but under different keys. Translated into Vietnamese sports terms: two people counting the same thing under two different definitions. One counts shots, the other counts shots on target. One defines a dangerous attack as one containing at least one pass into the box, the other defines it more broadly. Two tables, both correct, both about the same match, never reconciling.

All three share one uncomfortable feature: none is the analyst's fault. He received a broken data package from someone else and was asked to produce conclusions from it. In that situation, the only move that preserves professional dignity is to state clearly that the package was broken.

Degraded-profile analysis: working with twenty percent of the data

The report did not stop at refusal. It proposed a working mode called degraded-profile analysis: analyse populated dimensions at full depth, mark empty dimensions as insufficient information, and tag every conclusion with its evidentiary base.

This is a method I had used for years without having a name for it.

Take Vietnamese table tennis. Suppose I need to assess a player before the SEA Games. I have no rally-length data, no distribution of points by serve type, no win rate in extended counter-loop exchanges. Those exist only at events covered by the International Table Tennis Federation's collection systems, and even there incompletely.

But I have other things. I have the list of authorised equipment coverings, I have the history of the switch from celluloid to the 40mm plastic ball, I have domestic head-to-head results, I have rankings, and I have video.

So I analyse what exists. For instance, I can discuss the effect of the ball change: the plastic ball spins slower, so heavy-spin styles lose relative advantage against close-to-the-table speed styles. That is a claim with a physical basis, verifiable by comparing the results of the same cohort of players before and after the change.

And for what I lack, I state it plainly: insufficient data to assess tactical adaptability in prolonged counter-loop exchanges.

The difference between those two sentences is large. The first is a conclusion that can be challenged and checked. The second is an admission of limits. A piece with only the first looks strong and is actually weak. A piece with both looks weak and is actually strong, because the reader knows which parts to trust.

Layering data is how I stay calm in a mad transfer window. Every player file I split into three layers: measured metrics, eye observation, and rumour from insiders. The three are never mixed in one sentence.

Confidence labels: what Vietnamese sports language lacks

The report attached a confidence label to each hypothesis about the failure, and an overall label to the whole assessment, noting that the label rested on direct inspection of the supplied data package rather than on inference about the underlying article.

It is a small convention that changes how you read entirely.

In Vietnamese sports journalism we have no such convention. A transfer story sourced from an agent and a transfer story sourced from a photo on social media are presented in the same tone, the same font size, the same page position. The reader has no tool to distinguish them beyond guessing from the outlet's reputation. And outlet reputation cannot distinguish one story from another within the same day.

I once proposed to a few editors that every piece carry a line stating its evidence level. The proposal was rejected for a very reasonable reason: readers do not care.

They may be right. But I noticed one detail: those same readers care a great deal about the bookmaker price on the national team's title chances. They care about probability, as long as it is packaged as a readable product. A low, medium or high confidence label is also such a product. It just has not been packaged yet.

The counterintuitive angle: an empty report is worth more than a full one

Here I want to push back a little.

The natural reaction to a nine-dimension report where every dimension says insufficient information is: failure. A data group spent days producing a document saying they know nothing.

But weighed against a full nine-dimension report where eight dimensions are filled with confident-sounding speculation, I choose the empty one. The reason is accounting: an empty report causes zero damage, a fabricated report causes damage that compounds over time. The invented number gets cited, enters another file, becomes a premise for a transfer decision, and three years later nobody can trace it back to its origin.

I have been on the receiving end of that compounding chain. Once we dropped a player from a shortlist because a defensive metric looked poor in a compiled report. Six months later the data provider admitted the metric had been miscalculated due to a position-mapping error: the player had been assigned the wrong role, so every action of his was graded against another role's standard. Nobody was fired. The report is still in the archive.

But I must also state the rest of the counterintuitive point, otherwise I am flattering myself.

The auditor's own blind spot

An input audit can become a shelter.

When you discover the data is insufficient, you have a perfect reason to issue no judgement at all. And if that reason repeats often enough, it becomes a profession: the profession of pointing out other people's errors.

In transfer work I have seen such committees. They meet four times, each time concluding the file is incomplete, and by the end of the window they have signed nobody. Their caution is not methodologically wrong, but it carries an opportunity cost that never appears in the minutes.

The hard part of this trade is distinguishing two nearly identical states: not yet enough data to conclude, and enough data to conclude that no conclusion should be drawn. The first is the collector's job. The second is the decision-maker's.

The report that night avoided the trap with a single line: if only a fragment survives, send it, and I will run a degraded-profile analysis. The writer did not use the gap to stop. He used it to define precisely the boundary of what could be said.

After seven years, I believe in the silence between two numbers. That silence is not a place to rest. It is the place where you measure how far apart the two numbers are, and why.

Signal for the next cycle

What I will watch next season is not which metric gets imported into the V-League next. I will watch whether any outlet dares to print the line insufficient data directly beneath a statistics table.

If one does, that is a small but real signal: the industry has begun to distinguish not knowing from pretending not to know. If none does, we will keep getting full pieces, twelve smooth paragraphs, and none of the people writing them will know what they filled the gaps with.

There is one hypothesis I have not been able to test: whether the number of empty reports rises as data groups professionalise. Because a mature data group is a group that knows how to refuse. And a group that knows how to refuse will produce fewer pieces, fewer numbers, fewer headlines.

Whether a sporting ecosystem will accept paying in quantity for credibility is the question I leave open, the way a player opens a side-spin serve and waits for the opponent to guess the direction.