When the Data Runs Empty: The Art of Refusing to Fabricate in Basketball Analysis

Opening There are moments in this profession that never make it to broadcast. The blank screen is one of them. February 7, 23:40. Three data sheets open side by side, waiting for an internal API to push the numbers for the game I had agreed to analyze. Nothing arrived. Not a single row. Only the two team names surfaced on the header bar; everything else was white space. Forty minutes until deadline. I could easily have done what many in this trade do under the same circumstances: write. Write about the away team's star form. Write about the home team's switching defense. Write about how the coach rotates his rotation in the fourth quarter. It all sounds reasonable. It is all a product of imagination. That night I did not write. I sat still, looked at the white space, and remembered why I once swore I would never fill the blank with numbers I did not have. That lesson was expensive. It began four years ago, at a tournament I would rather not recall: the 2026 World Cup. Context: data is not a single block of stone Basketball was born a sport of numbers. Unlike football — where a midfielder can run off the ball for thirty minutes with no one recording it — basketball has had a box score since its earliest days. Every possession is a binary event: in, or out. Scored, or not. That binary nature makes basketball fertile ground for advanced metrics. A single NBA game today generates hundreds of variables: TS% (true shooting), OffRtg and DefRtg (offensive and defensive efficiency per hundred possessions), USG% (usage rate), EPM (estimated plus-minus), and dozens of derivatives. At the top of that chain sit weighted claims: «This player is the fifth most efficient shooter in the league.» Without a model, nobody dares say that sentence. But basketball data is not a single block of stone you can carve at will. It is a supply chain. It begins with the charting staff working the arena — the person deciding whether a pass is an assist, whether a deflection is a steal or an out-of-bounds turnover. It passes through data entry, cross-checking, and modeling. And it ends with the reader, the believer. In Vietnam, that chain is far shorter. The bulk of data about the NBA, the EuroLeague, or international competitions reaches Vietnamese audiences through multiple layers of retranslation — from English, sometimes through another intermediary layer, sometimes reduced to a screenshot with no date, no source, no original author. Each time data passes through another pair of hands, it loses a little context. Sometimes it loses its provenance entirely. And here is the point I want to make: when provenance disappears, the reader still feels the confidence of the prose, but has lost the ability to verify. Confidence replaces evidence. That is the ideal environment for fabrication. Based on my experience following games and transfer reports across many seasons, I have noticed a troubling pattern: whenever provenance disappears, the confidence of the writing rises rather than falls. People do not become more cautious when evidence is scarce. They write more forcefully. The core The blank space and the instinct to fill it What happened to me that night is a phenomenon researchers call confabulation — fabrication. Put an analytical system in front of too small a piece of information and it tends to produce a complete story: full of teams, players, transactions, metrics. The story sounds so plausible that the reader has no reason to doubt it. People are no different from machines. Faced with a blank space, the brain immediately fills it in. It cannot tolerate emptiness. It does not want to say «I don't know.» Look at how a transfer rumor spreads. It starts with an unsourced post: «Hearing Team A is eyeing Player B.» Within twenty-four hours, that line becomes an analysis of how Player B fits Team A's system. A day later, it becomes a piece about tactical consequences. Three days later, someone has drawn up next season's starting lineup. Not one link in that chain has evidence. Each writer grows more confident than the last, because each sees their own «source» — and that source is merely the previous writer, also fabricating. The fabrication is replicated. By the end of the chain, no one remembers that the origin was, in fact, a blank space. I have stood in the middle of such a chain. Not as an observer, but as the fourth link. I wrote an analysis of a transaction that, it turned out, no one had ever confirmed beyond a cropped image. My piece was clean, logical, and full of numbers. It was also entirely baseless. Who chose this number? This is the question I always ask of any metric, and the question I see least often on forums. When you read that a player shoots 42% from three, you are looking at a number. But numbers do not appear on their own. Someone chose it. Someone chose how many games the sample covers. Someone chose which time window to isolate. Someone chose whether to include or exclude heaves at the end of a quarter. Someone chose where to place shots that were fouled. The numbers do not lie, but the people who choose them do. I learned this the hard way. In November 2026, I was invited to write a column before a match I remember vividly. Based on a model combining four years of qualifying data, I declared a result with 94% probability. I wrote as if reading an already-issued verdict. The result went entirely the other way, and my column was mocked across forums. I once thought I was right. Qatar taught me I was wrong. What I missed was not in the model. It was in the 34°C heat and the air pressure that loosened the thigh muscles of players accustomed to playing at low altitude. Those variables were never in my spreadsheet, not because they did not exist, but because I had treated them as noise. I trimmed the data to fit my frame. I did exactly what every analyst is tempted to do: build the frame first, then stuff the data into it. Afterward, I set myself a discipline. Before any conclusion, I must check at least five foundational metrics: PPDA (the pressure applied to the ball carrier), xG chain, pass progression, tempo, and an independent defensive metric. If one metric says one thing and four say another, I am not allowed to write a conclusion. I must write about the contradiction. Who is actually present in the story? Basketball analysis usually begins by identifying characters. Who scored? Who made the decisive pass? Who defended without credit? But before you can identify the characters, you must answer another question: who is actually present? In a transfer equation there are at least two parties. Three, four, five if you count agents, families, and the old club. Each has its own motive to leak. An agent wants to raise his client's contract value. A club wants leverage in negotiations. A coach wants to send a message to ownership. None of them leak out of kindness to the audience. When a paper reports «Team X has reached an agreement,» the right question is not «what does this mean on the court,» but «who benefits if I believe this.» What is the reporter's relationship to the agent? Does the posting time align with a negotiation milestone? Is a third club being forced to react? A transfer is not a calculation; it is a negotiation between people and numbers. I once watched a player's valuation shift because of a run of articles within a single week. When I traced it back, the whole chain originated from a single source, and that source had a direct relationship with the club trying to sell. The number never changed. Only the story around it changed. And the market believed the story. The timestamp — the most neglected thing If there is one small detail most often cut when a story is retold, it is the date. A number without a date is a number that cannot be placed anywhere. It does not tell us whether it is fresh or stale. It does not tell us whether the player has changed, whether the injury has healed, whether the tactical system has shifted. Remove the date, and a stat from last season can sound like a stat from this one. In transfer rumors, the timestamp decides meaning. A tweet posted at 2 a.m. just before the deadline carries entirely different weight from one posted three weeks earlier. Same words. Same number. Two different truths. For Vietnamese audiences — who often encounter news through translations and screenshots — the loss of timestamps happens almost by default. A stat from three seasons ago can float online for years, reposted with no verification, gradually becoming a «fact» no longer tied to any moment in time. The discipline of refusal There are two kinds of failure in analysis. The first is concluding too early, before there is enough data. The second is concluding even though there is no data. The second is more dangerous, because it leaves no trace. No one knows the analyst fabricated, because the piece still reads fluently. In system design, the two options are called fail-open and fail-closed. Fail-open means continuing to run on weakened assumptions. Fail-closed means stopping when the input is missing. In generative analysis, fail-open is the primary driver of fabrication. The most honest systems are not the ones that always give an answer. They are the ones that can say «I have no data.» People are the same. A good analyst is not someone who always has an opinion. It is someone who knows when to stay silent. On the night of February 7, I chose silence. I sent the desk one line: «The match data feed has not arrived; I cannot analyze.» They were not pleased. But I kept the only thing worth keeping in this trade: honesty toward the blank space. The counterintuitive angle There is a counterintuitive way to look at all of this: the blank space is not the enemy. It is the best teacher an analyst can have. We live in an age where basketball data is so abundant that it becomes noise. Anyone can look up numbers, cite numbers, impress with big numbers. But precisely for that reason, value is not in having data. Value is in knowing which data to trust. The blank space is where an analyst's true nature is exposed. Faced with an empty sheet, the fabricator will fabricate and the honest person will tell the truth. There is no data to hide behind. No model to cover for you. Only one simple and difficult choice: say «I don't know,» or fabricate. In a market like Vietnam, where most data arrives through multiple intermediary layers, I believe admitting uncertainty has higher commercial value than pretending to be certain. A paper that says «we do not have the data to conclude» builds more durable trust than one that cites a number no one can verify. Trust, in the end, is also a metric — it just never appears in a box score. Data is a mirror; do not get angry when it reflects an ugly truth. A second counterintuitive angle concerns my own profession. The analytics community often treats new metric sets as a sign of progress. But new metric sets are not born in offices, they are born in crises. The Empty Arena Index that my team and I built in 2026 came not because we were clever, but because football had stopped and we were forced to find a way to reread what remained. Crisis teaches you to read data differently. Abundance teaches nothing at all. What I took from 2026 was not that I needed to be better at modeling. It was that I needed to be more honest about what the model could not see. When the arena empties, only data whispers the truth — but only to those patient enough to listen, and brave enough to hear even what they would rather not. Progressive conclusion The war in this profession is not between those who have data and those who don't. It is between those who can say «I don't know» and those who need an answer at any cost. When the next quarter begins and the data feeds pour in again, I will once more race the deadline, decode numbers, analyze. But I will try to remember that the most important question is not what the number is saying. The most important question is: if tomorrow the supply chain breaks again, will I write — or will I stay silent? And if you are a reader, remember this: the emptiness behind a number does not always reveal itself while you read. You only know it exists when you start asking — who chose this number, and what did they want it to say.