The Empty Dataset and Nine Analytical Dimensions: Where Basketball Data Journalism Most Easily Fools Itself

On Tuesday morning I opened the aggregated data file from the previous day's session and found exactly one line: 0. No league name, no team name, no metric column filled in. Beside it sat the nine-dimension framework I built in 2026 — tactics, player data, salary structure, league landscape, rules, locker room, risk, media, industry ripple. Four hours to deadline. The temptation arrived on the third second. Basketball carries such dense narrative traffic that a writer can produce a perfectly plausible piece about any team without verifying a single number. Pick a star, attach a “shifting role” to him, add a coach under pressure, close with a line about character. The piece runs 1,500 words, gets shared, and is wrong at the root. I took the other road. And what is worth noting is that the other road produced more information than a full analysis would have. The architecture I use has two tiers. Tier one extracts: league, team, player names, discrete factual assertions, source, publication date. Tier two analyses: it places those extracted fragments into nine professional dimensions to find the gap between expectation and reality. The hard rule of this architecture is simple — tier two is not permitted to manufacture its own raw material for tier one. That morning, tier one returned a structurally valid object with empty content. No error was raised. This is the most dangerous failure mode in data work: the system does not crash, it goes quiet. And silence, in any data pipeline, can be misread as “nothing is wrong”. My work in Hanoi sits inside a media market where most transfer content is aggregation: one account posts, ten sites repost, and by the eleventh site the phrase “according to sources close to the situation” has become fact. In that current, the most valuable thing a data journalist can hold onto is not speed. It is the ability to tier sources. I learned this fairly late. In 2026, still a data editor, I wrote that Hanoi FC deserved to win 3-1 rather than scrape a lucky 1-0 against Quang Nam, based on xG of 2.87 against 0.45, 68% possession and 14 shots inside the box. The piece was mocked as “football is not mathematics”. A week later, coach Chu Dinh Nghiem said he had rewatched the tape and adjusted his tactics on the strength of that analysis. That night, the press called them soulless. xG said otherwise, and I chose to trust xG. Dimension one — tactics — cannot be graded without a named system. “Solid defence” is a feeling, not a system. To argue, you need specific terms: drop coverage in the pick-and-roll, five-out, Spain pick-and-roll, hedge and recover. A team playing drop coverage and conceding pull-up threes at high efficiency is a problem measurable through the opponent's eFG% on that shot group. No named system, no shot group, no eFG% — and every judgement collapses into a description of the viewer's feeling. Even with a named system, playoff transferability is a separate question. A team that runs and rotates constantly in the regular season can survive on depth. In the playoffs the rotation shrinks to eight, every half-court possession is dragged into a phone booth, and death arrives from the place the regular season never forced anyone to solve: who scores when the system breaks. Dimension two — player data — is where self-deception comes easiest. Points per game is the most contaminated metric in the sport. A player averaging 18 a night on a team that loses 55 games, logging 12 fourth-quarter minutes at a 25-point deficit, is producing what the trade calls empty stats: statistical output accumulated in low-leverage context. TS% and eFG% are better than points, but still need a usage correction: the same 56% TS% means something entirely different for a player consuming 32% of possessions versus one consuming 18%. Then comes the test I consider most important: sample shrinkage. A player shooting 41% from deep over 30 games may be a 34% shooter in truth. The three-point line is the highest-variance indicator in the sport, and a 12-game hot streak is enough to generate an 80-million-dollar contract. The young-price bubble in basketball does not burst because teams are stupid; it bursts because small samples look like talent. Dimension three — salary structure — demands something far drier: a number. A rookie contributing 14 points and 7 rebounds a night on four million dollars is a strategic asset. The same player at 32 million is a problem to solve before next season. A player's value does not live in his legs; it lives in the ratio between output and the figure on the payroll sheet. That is why I always ask about the apron — the first and second thresholds — before I ask about tactics. A contract is only truly right when the number signs alongside the signature. Dimension four — league landscape — has three mandatory inputs: the age of the core, the years remaining on that core's contracts, and cap flexibility. Remove any one and every judgement about a “contention window” becomes belief. A team with a 24-year-old core, three years of control and 30 million in flexibility is opening a window. A team with a 31-year-old core, two years of control and a second-apron hard stop is closing one without knowing it. Dimension five — rules — is the most neglected dimension in Vietnamese sports journalism, despite deciding more than injuries do. Load management policy, supermax eligibility, apron penalties: each clause generates a behavioural model that every team optimises around. Without a named player and a named transaction, the dimension cannot be scored. And I refuse to default it to “low risk” — an unexamined rule surface is not a clean rule surface. Dimension six — locker room — is where the data is quietest and where writers fabricate most readily. Without a named head coach and without reporting on internal relationships, silence is not evidence of harmony. It is only silence. I once wrote a very tidy piece about a team that was “stable” purely because no bad news had leaked. Three weeks later, that team changed coaches. Dimension seven — risk — holds the most sophisticated trap. A risk matrix where every cell reads “insufficient information” will be skimmed by readers as “no risk”. That is more dangerous than having no matrix at all. The only defence is explicit labelling: not assessed, as distinct from assessed and found clean. Dimension eight — media — is the dimension I believe carries most of the analytical value. In transfer content during the rumour phase, roughly 80% of the value sits in the source tier, not in the content of the claim. A report that “team A is interested in player B” may be genuine insider information, or a line constructed to move a price. Identical content. Entirely different source tier. And when the source field is left blank, every credibility verdict is arbitrary. Dimension nine — industry ripple — is the most beautiful to draw and the easiest to render worthless. An impact map running from footwear to broadcast rights to regional markets to derivative markets only means something when a specific commercial signal exists. Without a signal, the map is decoration — and worse, persuasive-looking decoration. Nine dimensions, and all of them returned the same result. The result was not “there is nothing to say”. It was “there is not yet anything to say”. Here sits a paradox I consider central to this trade. Sports analysis does not lack data. It lacks people willing to say “there is not enough data”. The social cost of that sentence is too high: it sounds like weakness, while inventing a confident-sounding thesis is rewarded with page views. My experience says otherwise. In 2026, while most colleagues picked Brazil or Germany, I wrote that Croatia would reach the World Cup final because their midfield covered an average of 112 kilometres per match, the highest in the tournament, and the trio of Modric, Rakitic and Brozovic held a PPDA of 8.2 — a suffocating press. Croatia did not reach the final through luck. They reached it because their legs did not stop. But the same dataset betrayed me four years later: in 2026 I predicted Germany would escape the group stage on the back of the tournament's highest accumulated xG, and Germany went out. Japan posted a PPDA of 6.8 across their matches against Germany and Spain — a metric outside the dataset I had assembled before the tournament. It took weeks to digest and three months to rebuild the system. In 2026, when the Bundesliga returned to empty stands, I bet that home advantage would fall from 54% to below 50%. League-wide home win rate dropped to 48.7%; Dortmund won only 3 of their last 8 home games. The model was right on prediction and wrong on recovery, because I had failed to anticipate differences in training-ground quality and squad psychology. When the stands emptied, my model collapsed. I knew I had forgotten the human factor. The lesson is not that xG is useless. The lesson is that numbers show a trend, not a prophecy. I do not believe in gut feeling. But I believe in what gut feeling confirms once the data has signed off on it. So when a dataset is empty, the correct act is not to fill it with intuition dressed in terminology. The correct act is to record that it is empty, and to record why. Numbers never need us to defend them. We need them so we stop lying to ourselves. The signal for the next cycle is concrete. Before testing a claim, check whether there is a claim to test. Before asking “is this team actually good”, ask “how many games, how many shots, and who says so”. In a season where hundreds of analyses are generated each week from files nobody checks, the question of sourcing has become a tactical question.