Four Pages of N/A: The Silent Failure Inside Sports Data

Four sheets of A4, printed double-sided, clipped at the corner. Each page is divided into sections with bold headings, and beneath each heading sits a three-column table: category, assessment, notes. There is an "Evidence" section, a "Risk Flags" list, a "Risk Matrix" table. The show producer placed it in front of me twelve minutes before air and said: "Give me a read on this."

I turned to page one. The assessment column read: "N/A — insufficient information." Page two was identical. Page three, page four, the same. I asked: which player? He didn't know. Which team? Didn't know either. Which league? The file said exactly one word: basketball.

I spent those twelve minutes counting. No original headline. No source. Not a single information point. Not a single resolvable name. And yet at the end of the document there was still a section titled "Core Judgment," complete with a one-to-five-star information value table and a "signals to keep tracking" list.

I refused to go on air. The producer was annoyed, because the graphics were already rendered.

That night I understood something about my trade: sports analysis has just produced a new kind of product — a report with no data but enough formatting to reach television. It isn't rare. Nobody has simply bothered to count it.

This month is transfer season, and transfer season is the breeding ground for exactly that document. Every day brings hundreds of reports, thousands of posts, dozens of analytical tables about the same transaction. Audiences don't lack information; they lack a filter. But a filter only works if there is something inside it to filter — and that is precisely where the modern pipeline breaks.

I built my own database by hand, so I know what it costs. In 2026, when MLS shut down for 118 days, I was locked in a Miami studio with empty stadiums. The thing that had carried me for twenty years — a delivery built on crowd noise — was suddenly useless. I fell back on old reflex: I rewatched all 400 MLS matches from 2026 to 2026 and built files on 215 players across 12 criteria. One page per match, four columns: on-court events, player decisions, observed metrics, personal judgment. That archive is how I spotted Nani's high-speed running distance dropping 32 percent and predicted his decline the following season.

Those four columns are why I dare speak on air. I'm not smarter than anyone; I just know which page each sentence of mine is standing on.

There is a cultural gap I can't skip over, because it explains why this error is so hard to catch. I grew up in Manila, where fans love basketball with their hearts and trust their gut over any spreadsheet. Then I moved to the United States and worked in rooms where people dissect every possession into metrics. Those two cultures look like opposites, but they share one blind spot: neither checks whether the data it is using actually exists. Manila trusts memory. Miami trusts the spreadsheet. Both can be fooled by a document that looks correctly formatted.

The modern system works differently from the way I did. It runs in two stages. Stage one extracts the source article into structured fields: title, source, publication date, information points, core viewpoints, named entities, time sensitivity, source quality. Stage two takes those fields and runs them through a nine-dimension framework: tactics, player data, salary and cap, league landscape, rules and governance, coaching and locker room, risk, media narrative, industry ripple.

The framework is good. I've used crude versions of it for eight years. The problem sits in one very small line: when stage one returns empty, stage two still runs.

In this specific case, the input integrity check returned zero information points, zero entities, zero arguments. The domain label read only "basketball." Stage one failed completely.

The correct response — and the document I was holding chose it — is to write "insufficient information" in every cell and halt the chain. But the remaining structure was fully intact: all nine dimensions present, every heading present, every table with its rows and columns, the "Core Judgment" section still with a blank to fill. Another system, or another person under the pressure of airtime, needs only one prompt to turn this into fluent, unsourced prose that looks exactly like the real thing.

A wrong report filled in from empty data has the same structure as a correct one. That is the most dangerous failure mode in any data system: a failure that makes no noise.

In basketball, noise always exists. If I misread a substitution, the next quarter shows it and I have to correct myself on air. If I misassign a metric, the official box score hits me before the game ends. A document has no next quarter. It sits in an archive, gets cited again, and becomes the source for the next round of analysis.

Picture the same thing happening to a playoff series. A decent tactical report has to state pace, effective field goal percentage, minutes distributed across lineup groups, and switch rate. Without those four, every sentence about "tactical adjustment" is mood description. And if the report cannot even name which team leads the series, it has crossed from weak analysis into misrepresentation.

I came to this lesson through two failures at two different levels.

In 2026, aged 43, I was commentating for a Miami sports station. Modern analytics sites were taking over the industry, and I — a former midfielder with 12 years at the top level — publicly rejected it on air: players are not dry numbers. It blew up during Atlanta United versus Orlando City, when I called Josef Martinez a lucky striker, despite 19 goals in 20 matches that season. A 27-year-old colleague answered with a chart: Martinez's expected goals rate was 0.85 per 90 minutes, the highest in MLS that year. I had no reply.

I once thought xG was meaningless, until it explained why we lost.

I started keeping a handwritten match log. Four columns. From then on, every analysis I wrote carried a mandatory section: gut feeling checked against evidence.

Then came 2026. Aged 44, I went to Russia to commentate the World Cup. In the pre-match show before the quarterfinal, I declared that coach Roberto Martinez's inverted-fullback scheme would collapse under Brazilian pressure, and predicted Brazil 2-0. On 6 July 2026 the result was Belgium 2-1, with Kevin De Bruyne scoring in the 31st minute from exactly the advanced inverted-fullback position. Thirty days later, I rewatched all seven Belgium matches.

Belgium 2026 taught me that a golden generation does not automatically produce victory.

Those two failures taught me two opposite things. The first: my eye was wrong, the data was right. The second: I trusted a tactical model without rewatching the tape, and I was wrong. I set a personal rule — no comment without reviewing the footage — and every piece I wrote afterward opened with "after rewatching the match tape," with exact minutes rather than vague impressions.

Now I have to add a second rule, for a new kind of error: no comment until I confirm the data actually exists.

It took me two weeks to trust data, but twenty years to understand it still isn't enough.

So what is the minimum threshold for an analysis to be allowed to exist? I propose three conditions, and they closely resemble the three columns of my notebook.

First: at least three concrete information points. In basketball that might be a starting lineup, an actual shooting percentage, a specific minute of a specific possession. Three events is the minimum needed to discuss a trend rather than a moment.

Second: at least one resolvable entity — a name you can call out. A player, a coach, a team, a general manager. The industry calls this entity resolution, and it is a prerequisite for nearly every downstream dimension. No name means no analysis, only prose.

Third: at least one number. A contract value, a metric, a date. A number anchors the sentence to reality, and it is also the easiest thing to verify.

Those three conditions are cheap. They don't require a bigger language model, deeper data, or a larger budget. They require one gate in the right place — and one person willing to press stop.

There is one diagnostic detail I consider the most important in the whole affair: the missing headline. The title survives almost every kind of extraction failure. If a parser retrieves the body but loses the headline, that is a parsing bug. If it loses the headline too, the fault most likely sits before the fetch: a paywall, a JavaScript-rendered page, or a source that was never text — a video, a podcast, an untranscribed broadcast segment.

Three causes, three different fixes, and all three sit upstream. No prompt at the analysis layer can repair a source that was never read.

That is why I propose four operational metrics any sports data desk should pin to the wall. Extraction completion rate: count information points per article in a batch. Entity resolution rate: the share of articles yielding at least one name. Source-field population: the share with a source and a specific date. And failure-mode clustering by content type: HTML, PDF, video, paywall.

These four metrics have a property I like: they catch systemic failure before it becomes published copy. If 5 percent of articles in a batch yield not a single name, you don't conclude the writing is bad. You inspect the pipeline. In basketball, if 5 percent of your possessions have nobody touching the ball, you don't yell at the players — you look up at the 24-second clock.

And this is where it touches transfer season, which you read about every day. A decent transfer report has to answer very concrete questions: how many years remain on the deal, is there a player option or a team option, what is the salary structure, where does the team sit relative to the cap and the tax aprons. Without those, the report can still be written beautifully, confidently, and be entirely unverifiable. It is an empty capsule with good typography.

I have seen enough of them this season to believe we are in the middle of a crisis that is very hard to see. The crisis is not a shortage of data. It is empty data presented in the language of full data.

The first reaction most people have to this story is to retreat to the expert eye. I understand that reflex, because I lived inside it for twenty years. But my eye was wrong twice, at two different levels: wrong about a player, and wrong about an entire tactical system. If the eye is evidence, my evidence used to be terrible.

Data is only a map; the game is the storm. A beautifully drawn blank map is still a blank map, and it is worse than no map at all, because it makes people believe they have already travelled.

The part few people in the industry want to say out loud: the system rewards confidence, not emptiness. A commentator who says "I don't have enough data to conclude" on air looks far weaker than one who says "I predict 2-0." I once chose the second option, and I paid with thirty days of tape review. The real cost wasn't that I was wrong; it was that I taught the audience that confidence is a form of evidence.

The risk here is asymmetric. A misread possession is corrected the next quarter. An empty document filled in wrongly lives in an archive, gets cited again, and becomes "historical data" for the next analysis. The error doesn't disappear; it multiplies. And because every analytical framework resembles every other, the wrong version looks identical to the right one at every formal checkpoint.

In the coming months, when a transfer report lands in front of you, a cheap test will answer quickly: who does it name, what number does it give, does it carry a date. If all three are empty, you are reading an empty capsule — enough format to believe, not enough substance to verify.

And for those of us who do this for a living, the harder question sits elsewhere: do we have the courage to say "I don't have the data" on live television, or will we keep choosing the number that sounds more reasonable?