When a Basketball Analysis Has Full Structure and Not a Single Fact

2:14 a.m. I open a nine-part analysis file.

All nine parts are there. Every part has a table. The risk matrix has a probability column, an impact column, a mitigation column. A one-to-five-star rating scale. Column headers lined up like a starting five waiting for the tip-off. Then I read the body.

First cell: N/A — insufficient information. Second cell: N/A. Tenth cell: N/A. By the fortieth cell I stop counting.

The framework is intact. The content is empty. No player is named. No team is identified. No statistic appears, not even a wrong one. No source. No timestamp. What remains is the shape of an analysis — and the shape is flawless.

The file was produced by a pipeline failure: empty input data, untouched analytical frame. But sitting with it long enough, I understood I was holding the most accurate portrait of the sports news industry I have lived inside for ten years. The only difference: this file is honest about its own emptiness. Most of what we read every morning is not.

That is why I am writing this.

Every decent basketball analysis runs through two layers. The extraction layer turns a source article into discrete information points: who said it, when, where, what was said, on what sourcing. The reasoning layer stands on top of those points. When extraction returns an empty list, reasoning has no floor.

A decent writer stops and says: I have nothing in hand.

The industry does not stop. Because the framework still looks good.

In 2026, seventeen years old, I spent 72 hours rewatching the final 14 possessions of Game 5 of the NBA Finals between the Cleveland Cavaliers and the Golden State Warriors. I built a hand-made spreadsheet from four stats sites, cross-referencing Kevin Love's eFG% — 38.5% across the series — against six possessions where he stretched the defense wide enough to generate 10 direct points for LeBron James. The piece ran 2,000 words and drew 47 reads.

The article was not saved. But it taught me the thing that remains the spine of the craft ten years later: data has to be built from the bottom up, one possession at a time. Built top-down, it is only decoration.

Mid-pandemic in 2026, with leagues suspended, I retreated into the archive and built the podcast "Coverage Zone." Nine weeks, eight Olympiacos games in the EuroLeague, measuring the average distance between the two guards in pick-and-roll coverage: 4.7 metres; and how they forced opponents to the right wing 63% of the time. Thirty episodes, 25 minutes each, one situation per episode.

That coverage zone taught me a tactical situation can only be read with three things present: who does it, when they do it, and against whom. Remove one, and the rest becomes anecdote.

Since then I have read hundreds of basketball pieces a week, most of them translations and rewrites. I started logging a simple metric: how many of them carried all four basic fields — source, timestamp, entity, information point. That ratio has fallen every season. In the busiest week of this season's trade market, I counted 11 out of 140.

The rest of this piece is about those four fields, and why they matter more than any well-turned sentence.

Source: the most-abandoned field

During trade season, anything can be rewritten into a sentence with no source. "According to US media, this star wants out." Syntactically correct, publishable, and informationally worthless.

In July 2026, an entire basketball summer revolved around the story that Damian Lillard wanted only Miami. There were analytical pieces. There were pre-built salary sheets. There were three-team scenarios. There were calculations of exactly how many contracts Miami would have to shed to match salary. On 27 September 2026, Lillard went to Milwaukee. That entire body of analysis — thousands of articles, millions of reads — had a predictive value of zero. The reason was not the writers' competence. It was that they had no source, only heat.

When the source field is blank, three completely different kinds of information look identical on the page: a reporter with real relationships inside an organisation, an unverified aggregator account, and an agent's staffer inflating a client's price. In Vietnamese, all three become "according to US media."

Those three have wildly different reliability. And the cost of telling them apart is close to zero: one outlet name, one attribution line, one link. The industry does not pay that cost, because speed sells and sourcing does not.

In a newsroom running on a 15-minute cycle, every rewrite is a trade-off decision. Checking sourcing takes 20 minutes. Rewriting takes 20 minutes. The time budget covers only one. The industry has chosen, consistently and without secrecy, and that choice shows up in exactly the number I have been counting: 11 out of 140.

Timestamp: the value of a rumour decays by the hour

A live rumour and a three-week-old rumour are indistinguishable without a timestamp. Their value differs by an order of magnitude.

In June 2026, the first lines about Kawhi Leonard wanting out of San Antonio appeared. Over several weeks, each new detail spawned another layer of analysis, each layer resting on the one before, none going back to check the timestamp of the original. On 18 July 2026, Toronto announced the deal: Leonard and Danny Green to the Raptors, DeMar DeRozan, Jakob Poeltl and a 2026 first-round pick to San Antonio.

The notable part lies elsewhere. The best analysis of that trade was written after it happened. The most-read pieces were written before, when nobody knew anything, and most of them picked the wrong destination. A rumour does not lose value because it is wrong. It loses value because it is old.

A complete timestamp — down to the hour, not just the day — turns a rumour into something checkable: when was it true, and has it been overtaken. In this industry, three weeks is a generation.

The name: one entity unlocks everything

In that nine-part file, there is one line I read over and over: name a single team or a single player, and five of the nine analytical dimensions become runnable immediately.

That holds true at the crudest level of this trade. With a player's name, I can open the stat database, check usage rate, compare actual shooting efficiency against expected, test whether the number is inflated by garbage time. With a team's name, I can open the cap sheet, see where they sit against the tax line, how many picks remain, which contracts expire, how long the contention window stays open.

Without a name, everything collapses into adjectives: excellent, concerning, explosive, in crisis.

A piece with no proper nouns still reads smoothly. It just cannot be verified. And a piece that cannot be verified cannot be wrong — which sounds like an advantage, until you notice it cannot be right either.

Information point: the atomic unit of evidence

This is the hardest part, and the sloppiest.

An information point is the smallest citable unit of fact, with a source, with a timestamp, able to stand alone without a sentence wrapped around it. In the 2026 Kevin Love piece, the information point is not "Love played poorly." It is three separate lines: eFG% of 38.5%; six possessions dragging opposing defenders beyond the three-point line; 10 direct LeBron James points generated from precisely those stretches.

Assemble the three and the story flips. Read only the first, and the story matches the existing prejudice. That is the whole problem.

The 2026 World Cup gave me a cleaner example. Mesut Özil was heavily criticised after Germany's group-stage exit. I calculated his expected-goals contribution in the two matches he started — the 0-1 loss to Mexico and the 0-2 loss to South Korea — and compared it with his Arsenal season. But the decisive detail sits in the third match: Özil was benched when Germany beat Sweden 2-1 through Toni Kroos's free kick in the 95th minute. The only match Germany won was the one he did not play.

The story that "Özil was the cause" does not survive that test. It survived a long time on television anyway, because television has no cell to fill in.

Euro 2026 taught me the reverse lesson, about good data rescuing a wrong prejudice. I measured the average distance between Italy's five defenders across 120 minutes against Belgium in the quarter-final: 4.2 metres, nearly a metre tighter than in the group stage. The number fit the "parking the bus" story perfectly. I called an Italian assistant coach I knew from a forum, and the conversation ran three hours. He said the compactness was a consequence of Belgium losing the ball in unwanted positions, not an intention. I had to rewrite the entire 3,500-word podcast episode. It became the most-listened episode of the month, past 5,000 plays.

The lesson sits here: a metric that fits an existing story is the most suspect metric of all. It is not wrong. It has simply not been tested by anyone with the standing to refute it.

Unknowns and zeroes

There is a subtler error in that nine-part file, and it is everywhere in the industry.

When an analytical dimension has no data, people tend to fill the blank with the label "low risk." That file refused. It wrote "unknown," and explained that unknown differs from low; that an unclear injury history is precisely the condition under which bad decisions get made.

Picture a contract evaluation where the option structure is unclear: whether the extension carries a player option or a team option. Those two scenarios are valued completely differently, and the gap decides an entire summer's personnel strategy. Writing "low risk" into a blank cell is a way of lying through silence.

This is why I always read the footnotes before the lede. Everyone is good at the lede.

Format as camouflage

This is the part that had me opening my laptop at 2 a.m.

A document with full headers, all five sections, bolded key insights, an open-ended closing line — will clear an editing process far more easily than a messy but honest set of notes. The framework creates the sensation that someone has done the work. That sensation spreads to the reader, and it spreads fast.

That nine-part file, if someone deleted every "N/A" and rewrote it in a confident voice, would become a formally valid analysis, publishable anywhere, and nobody would notice it contains nothing. It is not wrong. It is empty. And emptiness has no rebuttal.

A complete framework is the winning machine of credibility. And the winning machine is only an illusion until someone is willing to break it.

The contrarian read: the danger is not fake news

The whole industry worries about fake news. I think the bigger danger is the piece that is correct in form and empty in substance.

A blatantly false claim dies in three seconds in the comment section. A 1,500-word piece with full structure, bolded takeaways, an open ending — that one lives. It gets shared. It gets quoted in tomorrow's piece. Three weeks later it becomes "according to media reports." By then nobody remembers where it started, because there was never a timestamp to remember.

Here I have to check myself. If this argument sounds too obvious, it is probably right because it is obvious. I will take that risk rather than force it into a thunderclap discovery.

One thing I want to be clear about, because it is easily misread. Fan emotion is valid data. When a team loses three straight and the arena starts booing, that is a real variable: it changes how the coach rotates, it changes the press conference, it changes the pressure on the front office. A decent writer does not discard that emotion. They place it in its proper cell — an information point, timestamped, sourced to the arena itself — and analyse it like any other fact.

What I refuse is using emotion in place of evidence, then labelling it analysis.

The Vietnamese layer

I write this as someone inside that ecosystem, not someone standing outside criticising it.

Most Vietnamese-language basketball content is rewritten from foreign sources. That process routinely strips exactly three things: the original outlet's name, the timestamp, and the specific proper noun. A report about an import player in the VBA can pass through four translations and emerge as "an import player is reportedly in negotiations." By the fourth pass, nobody knows who the original source was, and nobody asks.

The problem is not the translator. It is that the pipeline has no valve: when the source field is blank, the process still lets the piece through. In a small enough newsroom, one person can be reporter, editor and publisher at once. There is no step at which to ask "where is the source." And a question nobody asks quietly disappears from the workflow.

What is worth saying is that Vietnamese readers are not more credulous than American ones. They simply have fewer tools to check. Someone reading in Saigon at 11 p.m. cannot pull up an NBA team's cap sheet to see whether the number in the article checks out. They have only the article, and the article has full formatting.

The variable for next time

Basketball never ends with the buzzer, it ends with a question. This week's question is specific: strip the formatting from the piece you just read — remove the headline, remove the bold, remove the closing line — and how many citable facts remain?

Four cells to check. Who said it. When they said it. Who it is about. And on what fact.

None of those four requires domain expertise. They require one habit: reading slower than the writing.

Every result is a deliberate lie. So is the news report, even when the writer did not mean it. And the summer of 2026 taught us that the pain of defeat is also a form of knowledge — provided we are willing to reread it through data rather than memory.