Nine Tabs Full of N/A: The Day the Data Room Decided to Write Nothing

At 9:12 on a Tuesday morning, I opened the spreadsheet in front of me and counted. Nine tabs. Nine tabs, and all nine returned the same value: N/A.

Tactics tab: N/A. Player metrics tab: N/A. Salary cap tab: N/A. Rulebook tab: N/A. Risk tab: N/A. Even the line labelled “Article Title” — the thing that should exist before the machine has read a single word — read N/A.

The only thing still alive in that spreadsheet was a label: basketball. A domain tag, with not one number attached to it.

In thirteen years in this trade, I have grown used to mornings when the spreadsheet returns ugly numbers. A team loses, a model misses, a metric falls. But a spreadsheet returning ugly numbers is still a spreadsheet with numbers. A spreadsheet returning N/A is a different kind of silence: it does not say “I was wrong”, it says “I have nothing to say”.

People assume a good analytics room is one that can reach a conclusion in every situation. A good analytics room is one that knows exactly when it is not yet permitted to conclude.

The story starts with a two-step process that nearly every professional sports data centre runs. Step one breaks the source article into information points: title, source, article type, a one-sentence summary, the author's stance, the article's purpose, and the list of entities mentioned. Step two uses those very points as the foundation for the deep report.

Step one is the cheapest step and the most overlooked. Nobody praises an extraction layer that works. Only when it breaks does anyone look underneath.

And it broke. Not a collapse — no red error banner, no warning, no phone call. It broke by returning a beautiful skeleton: every field name present, every label present, every instructional note present, and not a single value inside. An empty mould.

To outsiders, this is a boring technical incident. To people in the trade, it is the most dangerous class of failure in the entire analytics chain — more dangerous than the machine producing a wrong number. A wrong number can at least be argued with. An empty mould invites people to fill it in.

This part deserves detail, because it is exactly where most Vietnamese sports data rooms currently stand, whether or not they have named it.

A skeleton with labels but no flesh is the signature of an extraction failure, not the signature of an article with no content. This is the single most important distinction in the whole story. A genuinely empty article cannot produce a domain label. To assign the label “basketball”, the system must have read something — a headline, a URL, a section, a tag. A surviving label means textual traces existed somewhere. The body vanished.

Cause one sits on the source side. Major sports outlets now render content in JavaScript, and a simple collector will receive a blank page. The machine raises no error, because to the machine a blank page is a valid answer.

Cause two sits at the access layer: a paywall, or a regional block. The machine receives a landing page, extracts a few labels, and stops.

There is a third possibility at the encoding layer. If the origin was Chinese or Japanese, a character-set fault in the middle layer can swallow the body whole while leaving the English field labels untouched.

All three paths arrive at the same place: the system has no idea it failed.

And this is where I want to linger, because it bears directly on Vietnamese basketball.

Over the past three seasons I have had occasion to look at the internal stat sheets of several clubs in the domestic professional system. None of them lacked numbers. What they lacked was a validation gate willing to say this batch is unusable. A single game can be hand-recorded by two different people, the two people write two different figures for the same possession, and both sheets get pushed into the system with nobody reconciling them. The error is not in the model. The error is that nobody is tasked with saying “hold on”.

I once sat in a meeting where a team's rebound figure was read out as 41, while the video showed the true number was 33. Those eight rebounds did not come from dishonesty. They came from someone in the room “guessing” the missing portion instead of leaving it blank. And because nobody knew which portion was a guess, the entire stat sheet lost its value.

That N/A report refused to do precisely this. It refused to guess.

A sheet full of N/A is, informationally speaking, worth more than a sheet full of numbers of unknown provenance. That sounds backwards, but it is the foundation of every serious data system. An N/A sheet tells you exactly one thing: do not decide based on me. A sheet full of numbers of unknown provenance tells you something far more dangerous: go ahead, use me.

The source report logged four warning signs, and all four belong on the wall of every sports analytics room.

Fabrication risk at the highest level. With an empty input, every conclusion produced is a product of imagination wearing an expert's jacket. It is many times more dangerous than a blank report, because it looks credible.

Loss of provenance. Both the title and the source were blank. Which means that even if the content is later recovered, nobody can prove where it came from. In this trade, a number without a source is worse than no number.

Silent pipeline regression. If the extractor breaks across an entire class of sources — paywalled outlets, JavaScript-rendered outlets, foreign-language outlets — the fault will recur undetected, because each time only one article is lost.

False confidence from one surviving label. The “basketball” tag was the only thing left, and precisely because it was left, a skimming reader will assume the report has content. This is the most subtle trap of the four.

These four signs reach beyond one system. They describe exactly how sports data fails silently at every level — from a hand-written sheet at the arena to a player-valuation model.

One line bears repeating, the one I still say to young coaches: numbers do not lie, but they do not tell stories either. An N/A sheet tells no story at all. The analyst's job is not to tell it on the sheet's behalf, but to state plainly: this part is empty, and here is why.

There is one detail in the source report I like most, because it is very human: the author noted that if only three things were resupplied — title, source, and at least three information points — all nine analytical dimensions would run in a single pass. Three things. No new model, no bigger dataset, no extra server. Just three fields filled in correctly.

In practice, such a gate needs only a few lines run before every report: empty information-point list, stop; empty title or source, stop; a domain label with no supporting text span, downgrade confidence to the lowest tier. Those lines save no model, but they save an entire room's credibility.

The hardest part of this trade is not the computation. The hardest part is making sure the input is not empty.

Everyone says data matters. The contrarian angle sits elsewhere: in sports, the person who says “I don't know” is paid less than the person who says “I think it's”.

A technical meeting runs like this. The coach asks: which wing does the opponent prefer to shoot threes from? The person who answers “I don't have the numbers yet, let me check” opens a gap in the room. The person who answers “left wing, I remember it that way” fills that gap instantly, and is usually credited with knowing the situation. Gaps in meeting rooms are always filled by what is available, not by what is correct.

That is why professional sports analytics processes, however well funded, still collapse at exactly one point: people are not rewarded for having stayed quiet.

That N/A report ran against the instinct of an entire industry. It stayed quiet, and it recorded the reason for staying quiet. In terms of professional discipline, that is far harder than a bold prediction. A bold prediction is right 50% of the time. Well-timed silence is right 100% of the time, and nobody hands out trophies for it.

Every coach talks about feel. I have no feel; I have standard deviation. But the standard deviation of an empty set has no value — it is not zero, it simply does not exist. And that is the whole problem.

I am not naive enough to tell every data room to simply stop. Some games do not permit stopping. In those cases the job is to state the assumptions, state the confidence level, and state the scenario in which the conclusion fails. Those three lines cost far less than the price of one decision built on invented numbers.

Before the next round, there is one small thing anyone working in sports data should do: open your own spreadsheet and count the empty cells. Do not fill them. Just know where they are.

An empty cell whose location you know is a harmless empty cell. An empty cell you filled from memory is no longer an empty cell — it is a number, and it will enter somebody's decision on some afternoon, and nobody will remember where it came from.

Data is a monastery: the less noise, the more clearly you hear something trying to speak. But a monastery also has empty rooms, and a good keeper knows which room is empty because nobody has arrived, and which is empty because someone swept it clean.