Nine Empty Boxes in the Video Room: The Discipline of Saying "Not Enough Data" in Basketball Analysis

The video room went dark at 1:47 in the morning. What stayed on the table was a forty-one page report, spiral bound, with a cover page, a table of contents, twelve charts and nine analytical sections. The head coach turned to page eleven, stopped, and read aloud: "Offensive system: insufficient information. Key player: unidentified. Salary structure: insufficient information. Coaching staff: unidentified." He folded the document, laid it face up, and said something I still remember word for word: "You have handed me a blank book."

I said yes. And it was the most honest report I have ever submitted.

That night our system ingested an empty file. The tracking cameras failed fourteen minutes into the first half. The extraction layer ran exactly as designed; there was simply nothing to extract. The next morning someone asked me the usual question: "So, is this team good or bad?" I said: "Don't know yet." Three words that cost far less than a wrong report, and irritate people a hundred times more.

Eighteen years in this trade have taught me that sport has one very specific fear: the fear of white space. An empty cell in a report reads as laziness, as unprofessionalism, as someone who does not deserve a seat in the video room. So people fill it. They fill it with instinct, with the memory of a game three months ago, with a metric lifted from a website of unknown provenance. Once the cell is filled, the report looks complete, the meeting runs smoothly, nobody objects.

Then the season ends and the club is exactly where it started.

How we build the frame, and how the frame indicts itself

Data consulting for a basketball club in Vietnam looks nothing like the popular image. There is no air-conditioned room with curved screens and numbers streaming like a Hollywood film. Most of the time is spent reconciling three sources for the same metric and discovering that all three trace back to one article by one reporter who was not at the game.

Our process has two layers. Layer one decomposes an article or a game into discrete units of fact: title, source, content type, information points, core viewpoints, named entities, time sensitivity, source quality. Layer two runs those units through nine analytical dimensions: tactics, player data, operations and salary cap, league landscape, rules and governance, coaching and locker room, risk, media narrative, and industry ripple.

That night, layer one returned a structurally valid and entirely empty frame. Title: none. Source: none. Information points: none. Entities: unidentified. Time sensitivity: not assessed. Source quality: not graded. The only populated field was the domain label: basketball.

Such a frame puts layer two in the hardest position there is. All nine dimensions have to answer the same question: with no input, judged on what?

Here is the point I want the reader to carry through this entire piece: an empty report in perfect format is more dangerous than an obviously broken one, because nobody can see the crack. When your data file opens with an error message, you know you have work to do. When it returns a blank table with all the column headers in place, you are tempted to sit down and start writing.

I almost wrote. I opened the document, typed the first line, and deleted it. What follows is an account of exactly what I found walking through each dimension on an empty input, and why every such walk teaches me more than a win does.

Dimension one: tactics without a subject

To evaluate an offensive system, the first requirement is knowing its name. Without a name, every analysis drifts. You cannot discuss the efficiency of a pick-and-roll scheme without knowing who sets the screen, who handles the ball, and whether the opposing defence chooses to switch or drop.

In modern basketball the gap between those two defensive choices is far wider than a box score suggests. A drop coverage team invites threes from above the arc. A team that switches everything accepts losing the matchup within two metres of the rim. Both can win, provided the personnel fit the choice. That is why I always test this dimension with a single question: who was this system designed for.

With no answer, every comparison table is decoration. We record four cells — sophistication, execution, personnel fit, key data — and all four return the same value. Cannot be assessed.

There is a very human temptation here. Facing an empty cell, my professional reflex is to fill it with a familiar example. I could reach for Denver and Nikola Jokić to illustrate a system built on a centre as the distribution hub. I could reach for San Antonio and Victor Wembanyama to talk about a wingspan that alters the geometry of the floor. But that swaps the subject. The reader would think we are analysing their games, when in fact we are reciting our own memories.

I call that borrowing a body. It is the most common reason a scouting report looks persuasive and predicts nothing.

Dimension two: a player with no name

The player data section is where I am tempted most. Four tiers sit ready: basic, efficiency, impact, usage. Each tier has three to five columns. A table that handsome looks like it is waiting to be filled.

It will not be filled.

Because no player is named, the entire validation methodology I consider the core value of this profession becomes meaningless. I cannot check a statistical line for empty stats — the kind of player who scores twenty a night across forty minutes while his team loses more than it wins whenever he is on the floor. I cannot check playoff shrinkage, when minutes tighten, opponents study harder, and everything pretty about the regular season gets peeled back layer by layer.

One of my most expensive lessons on this came from football, and I still tell it to every new intern. In June 2026 I was twenty-five, an assistant analyst at a sports outlet in Hai Phong. During the World Cup group match between Switzerland and Serbia I counted Granit Xhaka touching the ball one hundred and twelve times, with only thirty-four per cent of his passes directed forward. I wrote a piece criticising an excessively safe style. Coach Petković replied that football is not mathematics. Three days later Switzerland won 2-1, including eight decisive forward passes I had not seen.

My error was not the number. My error was looking at only one number. I ignored the metric measuring the opponent's pressing intensity, where Serbia ranked near the bottom. Xhaka's sideways passes were a consequence of having nobody ahead of him, not evidence of timidity. The same number, two opposite stories, and I picked wrong because I was lazy.

Since then I hold myself to a rule: before publishing, check at least five underlying metrics. No five metrics, no article.

Numbers do not lie, but the person choosing them does.

Back to the empty player table. What I refuse to do is explicit: attach a real player's statistics to a report in which no player exists. Doing so creates a false evidentiary chain between input and output, and that false chain outlives any single article. People will cite it. People will use it to sign someone. Nobody will check the source.

Dimension three: the cap, the hungriest dimension

Of the nine, this one consumes the most data and lies the easiest. You need a cap sheet, luxury tax thresholds, both apron levels, a pick ledger, and an absolute as-of date. Miss any one of those and a single transaction can overturn your conclusion.

Professional basketball is the sport where salary structure changes tactical substance. After the 2026 collective bargaining agreement took effect in the 2026-24 season, the second apron turned keeping three max-level stars into a question of rights, not just money. A team above that line loses access to the mid-level exception, loses the ability to aggregate salaries in a sign-and-trade, loses first-round picks at the end of the first round in the near future. None of that appears in a box score, and all of it determines your roster three seasons out.

A club in Vietnam once asked me why we do not simply import that formula into the domestic league. I told them it is like measuring height with a scale. The Vietnamese professional game runs on a different domestic-and-import player mechanism, incomplete public budgets, and far shorter contract horizons. Applying the apron of a billion-dollar league to a structurally different competition is the most common import error in my profession.

That is the fourth item on the list of errors I keep for myself: applying American data standards to a Vietnamese context without checking provenance, collection method, and the year of the data.

Dimension four: which league, still unknown

The domain label says basketball. That is all. But basketball includes the NBA, FIBA, EuroLeague, the Chinese league, the NCAA, and Vietnam's professional competition. Each runs on different rules, calendars, transfer mechanisms and competitive incentives. The NBA has a draft lottery that creates an incentive to lose. EuroLeague has no draft. FIBA imposes far stricter naturalisation conditions.

Identifying the league is step one of any re-run. Without it, roughly half the downstream framework must be swapped for another rulebook. This matters enough that it should precede tactical analysis, because tactics in Europe and in the United States are measured with two incompatible metric systems.

League landscape is also the one dimension where I can state something certain even on an empty input: every tiering model has an expiry date. A team placed in the title-contender bracket in October can sit outside the play-in by February. A standings table is a photograph, not a map.

Dimension five: rules only answer when someone breaks them

Governance analysis is reactive. It needs a triggering event: a fine, a tampering allegation, an extension dispute, a signal from collective bargaining. No event, no analysis.

In basketball this dimension usually appears in two very different forms. The first is fines and suspensions, where the real question is proportionality between act and penalty. The second is load management, where the league has issued a policy requiring star players to appear in nationally televised games, with exceptions for age and injury.

Both forms need a name. Without one, I cannot say whether a club is exploiting a loophole cleverly or shooting itself in the foot.

Dimension six: the locker room, the easiest place to fabricate

This is the dimension that worries me most in every analysis, because it is the only one where source quality determines almost the entire value of the conclusion. Locker room reporting runs across an enormous range: from a beat reporter with direct relationships to an assistant coach, to a social account summarising what someone heard from someone else.

In this analysis, source quality was never graded at layer one. Which means that even if the input had contained content, I would not know how much of it to believe.

I learned this the uncomfortable way. In 2026 I wrote a piece about the relationship between a head coach and a key player, based on three sources that all cited a single forum post. The article was coherent, sourced, logical. Six weeks later all three sources were deleted, and the original post turned out to be the work of an angry fan. I had written a complete analysis on top of a hypothesis that did not exist.

Since then I attach a source tag at layer one and refuse to let layer two infer it. Four tags: insider, beat reporter, aggregator, self-media. Each carries a different weight, and that weight must be written down before the first sentence is drafted.

Dimension seven: the one risk still alive

The six conventional basketball risk categories were all unassessable: competitive, contractual, personnel, rules, public opinion, systemic. But one risk remained alive, and it is more serious than the other six combined.

The silent propagation of an empty input into an output that looks full.

In a multi-layer analysis chain this is the most dangerous failure mode, because it makes no noise. The lower layer still complies with the format, still has headings, still has tables, still has conclusions. The reader sees a finished document. Only the author knows there is nothing inside.

The fix is technically trivial and culturally difficult: install a hard gate upstream. If information points fall below a minimum threshold, or the entity list is empty, the system must refuse to run layer two and fail loudly. Failure should be audible. In my trade, a system that fails silently is worse than no system at all.

Dimension eight: no headline, no outlet, no cycle

Narrative analysis needs two minimum inputs: a headline and a newsroom. Both were empty.

A missing headline is the strongest single indicator that extraction failed at the very first step, since headlines are usually the easiest thing to capture on a news page. A crawler returning a null title has almost certainly returned a null body.

Without a publication date you cannot place a story in its heat cycle. Every sports story passes through four phases: budding, peak, backlash, settling. The same information carries predictive value in the first phase and only archival value in the last. No date, no phase.

February 2026 is the example I use when training my team. A single transaction moving Luka Dončić from Dallas to Los Angeles forced nearly every player-valuation model in every newsroom to be rewritten within forty-eight hours. Those models did not get the numbers wrong. They got wrong the assumption that every front-office decision follows value maximisation. A story like that goes from budding to peak in under a day, and every analysis written in the second phase was already stale at publication.

That is why I stamp both the original publication date and the analysis date on every report. The distance between those two numbers is the shelf life.

Dimension nine: ripple effects with no stone thrown

Ripple analysis needs a triggering event: a signing, a trade, a broadcast rights deal, an international result. No event, no wave.

In basketball the chain runs through three tiers. Upstream is youth development and the agency ecosystem. Midstream is clubs, leagues, events. Downstream is broadcast, footwear and equipment, derivative markets. A star changing teams moves ticket prices, broadcast schedules, jersey sales, and the league's rights value in the following cycle.

In Vietnam the downstream tier is much thinner than imported models assume. A domestic league can grow its audience very quickly while equipment revenue barely moves, because the sports consumer market has not caught up with the speed of content distribution. Applying the ripple formula of a billion-dollar league to an emerging one is the second import error I have to remind myself about every week.

The reward for confidence, and the price of being right

Now for something uncomfortable about this profession.

A report full of white space does not sell. Nobody pays for a forty-one page document saying we do not know anything yet. Meanwhile a confident analysis, with numbers, charts and a clear conclusion, gets shared, cited, and invited onto television. The incentive structure of sports media rewards decisiveness, not accuracy.

I fell into that trap in the most expensive way. In November 2026 I was invited to write a column before the World Cup match between Saudi Arabia and Argentina. My model, built on four years of qualifying data, produced a very tidy conclusion: Argentina would win with a very high probability and a margin of at least three goals. I wrote it. The result was Saudi Arabia 2-1, built on a high defensive line and an offside trap that repeatedly caught Argentina's front line in the first half.

I once thought I was right. Qatar taught me I was wrong.

My error was not the model. My error was failing to put into the model the variables absent from the dataset: extreme heat, humidity, air pressure, and the fact that a team accustomed to playing at altitude had to operate unfamiliar thigh muscles in unfamiliar conditions. For two weeks afterwards I rewatched forty-seven matches from Gulf tournaments across ten years, and learned more from those forty-seven games than from four years of qualifying data.

Since then every pre-match analysis of mine carries a climate and geography section. And every projection carries a confidence interval. Readers noticed I had stopped writing in absolutes, and the paradox is that my credibility rose rather than fell.

But there is a deeper point that took me another two years to understand. The problem in Qatar was not that I lacked data. The problem was that my report had no empty cell in which to record that I lacked data. My model was designed to always return a number. A model with no exit for ignorance will always return a number, even when that number is meaningless.

That is the entire reason the empty report that night, despite infuriating the head coach, is the product I am proudest of this year.

Every number is a confession, if we are patient enough to listen.

There is one more layer I have not touched, and it is the part I want to leave with Vietnamese readers.

When we built the empty-stadium metric set during the pandemic shutdown, the group had four people and two hundred matches from two European leagues. We measured that central midfielders' running distance fell by nearly ten per cent in the first month after the restart, while line-breaking passes rose by more than thirteen per cent. The board was sceptical. We still persuaded them to sign a Brazilian midfielder on that model. After ten rounds he had four goals and three assists, including a counterattack the model had called correctly. The club climbed six places.

New metric sets are not born in offices. They are born in crises.

That lesson applies directly to Vietnamese basketball. Here we do not lack clubs, young audiences, or content. We lack measurement infrastructure. Many games produce only a basic box score: no positional data, no touch data, no defensive data. Under those conditions an analyst has two options: import metrics from another league and pretend they transfer, or build a coarser metric set that measures what you can actually see.

I choose the second, and the second always begins by admitting how little you can measure at all.

When the court is empty, only the data whispers the truth.

One assumption deserves to be stated plainly, because the discipline of admitting uncertainty is only worth anything when written as a sentence. My assumption is this: if the collection layer is re-run against the original source and returns a title plus at least three information points, most of the "cannot be assessed" verdicts above will convert into substantive conclusions, and the direction of the analysis could reverse entirely. I could be wrong on dimension one. The team I assumed had no clear offensive system may be running a very specific scheme that a broken data feed concealed. I do not rule that out, and I record it before writing the conclusion, not after.

At the same time, one thing I will not concede: no input, no verdict. That is not caution. That is the condition under which this profession exists.

Signals for the next cycle

Four signals I will track, and how to observe them.

First, the fill rate of layer one. If a re-run returns a non-null title and three or more information points, the full nine-dimension frame can run. This is the easiest signal to observe and the most valuable, because it separates a source problem from a process problem.

Second, league identification. A single team name or player name instantly gives dimensions three, four and five a subject, and moves roughly half the framework from substitution-required to applicable-as-written.

Third, the source-quality tag. If layer one grades it, the two most fabrication-prone dimensions — locker room and narrative — gain explicit weight instead of being written by feel.

Fourth, the publication date. One absolute date in layer one allows the story to be placed in the correct heat-cycle phase and allows the shelf life of every accompanying number to be calculated.

None of these signals is glamorous. None generates a headline. But they are what separates an analysis room from a re-publishing desk.

And if you ask why I am still in this trade after eighteen years, after one wrong call in Qatar, after countless occasions when data proved me wrong, the answer lies elsewhere. This profession does not grant me the right to speak the future. It grants me the right to say that today I do not know, that tomorrow I will know a little more, and that the day after I may have to delete everything I just wrote.

A report with nine empty boxes is not a failed report. It is a reminder that every filled box in the world was once empty, and that whoever filled it may have filled it with fact, or with something that merely sounded like fact.

The difference between those two people only becomes visible when the season ends.