The Blank Sheet in the Film Room: Why the Best Analyst Is the One Who Says 'Insufficient Information'

At two in the morning on March 14, 2026, I sat in front of a spreadsheet with 217 empty cells. The monitor light fell across the desk; beside it stood a coffee that had gone cold hours earlier. Outside the window, Hai Phong was quiet enough that I could hear the ceiling fan turn.

It was a tracking file for a Vietnamese professional basketball game, sent back by a contributor after the final buzzer. The box score was complete. Points, rebounds, assists, personal fouls — all present. But the column for "shot origin location" was empty in 217 cells. The column for "pressure timing on the ball handler" was empty entirely. The column for "who made the last off-ball cut in the fast break" had only nine rows filled, and all nine named the same person: the scorer.

I had enough material to write something that would sound highly professional. "Team X controlled the tempo but collapsed in the fourth quarter due to a lack of roster depth." Readers would believe it. Nobody opens the raw file to check those 217 empty cells. But I did not write it, and that moment — not any game — is where this piece begins.

The numbers do not lie, but the person who picks the numbers does. And here the person picking was nobody else. It was me, if I had filled those gaps with inference and then presented that inference as measurement.

Three kinds of gaps, three completely different stories

The first gap is the one never measured. In most domestic basketball leagues, a game has only two statisticians, sometimes university volunteers, and they must track both teams across four quarters. Scoring and personal fouls always take priority, because that is what the scoreboard needs to function. Who moved where to open a gap for a corner three — nobody pays for that to be recorded, so it does not exist in the record. This gap is harmless in the sense that it deceives no one. It is simply darkness.

The second gap is the one measured incorrectly. In 2026 I received a file with every column complete, so clean that I nearly fed it straight into a model. But when I checked it against the video, I found the recorder had calculated "pass count" by counting both teams' passes and dividing by two. The numbers looked plausible, and precisely because they looked plausible they were more dangerous than an empty cell. An empty cell forces humility. A wrong number that is neatly rounded does not — it invites you to build on top of it.

The third gap is the one removed on purpose. In 2026 I received back an internal dataset that had passed through three rounds of edits. Comparing it against my own archived original, 43 turnover possessions had disappeared. Not deleted for being wrong. Deleted because nobody wanted the player under evaluation to be seen in those possessions. This is the worst kind of gap, because it is not darkness — it is false light.

What I learned after years is not "do not trust data." It is this: a data gap is not a flaw in the process; it is the fingerprint of the process. The shape of what is missing — whether the holes are perfectly uniform, ragged, or absent according to a rule too perfect to be natural — tells you who collected it, what they were told to prioritize, and what they feared.

Three years ago I brought this argument to a club and was answered with: "Are you analyzing data or analyzing people?" I said both, because at the scale of our leagues the two cannot be separated. When a league draws a few thousand spectators a game and operates on a limited budget, data is not generated automatically the way it is in a major league. It is generated by a specific person, on a specific night, with a specific list of priorities. If you cannot read that priority list, you are only reading back your own bias.

The lesson from 200 games with no crowd

In June 2026, global football and basketball paused. Two colleagues and I at the club started a project nobody then considered serious: collecting data from games played after the restart, when the stands were empty. We gathered data from 200 matches across the Portuguese and Danish league systems.

The first result made the board suspicious: central midfielders' running distance fell 9.7 percent in the first month. That number ran against ordinary expectation, since everyone assumed players would run more without the pressure of a crowd. But the second metric is the one I want to discuss: line-breaking passes rose 13.2 percent.

Those two numbers only mean something placed side by side. Players ran less but passed more adventurously. That means they did not lose energy — they shifted energy from off-ball running to faster decision-making. But I have to state clearly what few are willing to say: that sample of 200 matches does not represent basketball, and does not represent all of football either. It represents one very specific circumstance — competition without spectators, inside a compressed window, on a congested schedule. Any extrapolation beyond that frame is inference, not conclusion.

When the arena is empty, only the data whispers the truth. But it whispers very quietly, and only about that empty arena. I wrote that limitation explicitly into the report I sent the board, and that is precisely what persuaded them more than the numbers themselves. A transfer recommendation that carries the warning "this model applies only to spectator-free competition" is more credible than one that simply says a player is good.

We signed a Brazilian midfielder on that model. After ten rounds he had scored four goals and assisted three — including one fast break the model had predicted in the right direction. The club climbed six places.

I tell this story not to boast. I tell it because it is often quoted wrongly. People remember "four goals, three assists" and forget "200 Portuguese and Danish matches," forget that we staked our credibility on an imperfect sample, and forget that if the player had torn a hamstring in round two the story would be told in the opposite direction. Success does not erase uncertainty. It only makes uncertainty easier to forget.

Switzerland, Serbia, and the mistake I do not like repeating

In June 2026 I was 25, an assistant analyst for a young sports outlet in Hai Phong. During the World Cup group match between Switzerland and Serbia, I found that Granit Xhaka had touched the ball 112 times but only 34 percent of those touches went forward. I wrote a piece attacking an excessively safe playing style, arguing Switzerland were tying themselves in knots.

Coach Petkovic responded that football is not mathematics. I took that as a dodge. Three days later Switzerland came back to win 2-1, with eight decisive passes in the second half.

I had ignored a metric sitting right next to the numbers I was staring at: PPDA, a measure of pressing intensity on the ball carrier. In that tournament Serbia ranked near the bottom for pressure applied. Which meant Switzerland had not chosen a safe style out of fear — they chose it because the opponent could not generate enough pressure to force them to gamble earlier. When they needed to, they raised the tempo and the opponent had no time to react.

Every number is a confession, if we are patient enough to listen. But I had heard one sentence and assumed it was the whole testimony.

From that day I set a hard rule: before publishing any conclusion, check at least five underlying metrics — PPDA, xG chain, pass progression, defensive shape, and touch share in the final third. Not because five numbers are truer than one. Because five numbers force me to slow down, and slowing down is the only thing that genuinely protects me from myself.

Qatar, and the time my own data refuted me

In November 2026 a major newspaper asked me for a column ahead of Saudi Arabia versus Argentina. My model then was the product of four years of qualifying data. It returned a 94 percent probability of an Argentina win and a minimum scoreline of 3-0.

I wrote exactly that. No confidence interval, no variable warning, not a single "if."

Saudi Arabia won 2-1. Their offside trap caught Argentina offside ten times in the match, mostly in the first half. My model contained no variable for that, because I had never included a variable I considered "outside football."

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

The variable I missed lived in no column I possessed: 34 degrees Celsius combined with high humidity, acting on the thigh muscles of players accustomed to cooler climates and different altitude. I then spent two weeks rewatching 47 matches from Gulf-region tournaments across ten years, only to log when teams lost their defensive structure in the second half. The result was not enough to build a model. It was only enough for me to understand I had been missing an entire dimension.

The biggest change was not adding a climate variable to every analysis. That is the most visible result and the least important one. The real change was how I present uncertainty. Since then, every forecast I make carries a confidence interval, and every analysis carries a passage stating clearly: if this assumption fails, the conclusion reverses at this point.

The paradox: the market pays for certainty, not for truth

This is the hardest part to say, and the most important.

An analyst who says "there is not enough information to conclude" will not be invited on air. An analyst who says "this team is certain to make the playoffs" gets quoted, shared, and remembered. The incentive structure of sports media rewards decisiveness, not accuracy. Sometimes those two coincide, but most of the time they do not.

This produces a consequence few name correctly: most fake data in the market is not created by bad people. It is created by honest people, under conditions that demand an answer before there is enough evidence. The recorder fills an empty cell because the boss needs the report on deadline. The analyst reaches a conclusion because the newsroom needs a headline before tip-off. Nobody in that chain thinks they are fabricating. Each person is only filling one small gap.

And here is the point I consider the genuinely counterintuitive angle of this piece: the solution does not lie in asking analysts to be more honest. We have plenty of calls for honesty already. The solution lies in changing the reward structure — making it acceptable to publish an "insufficient information" verdict without it counting as failure, and treating the retraction of a wrong forecast as a professional milestone rather than a stain.

In Vietnam this is harder than elsewhere, because the data infrastructure is thinner. A club abroad may have dozens of tracking cameras and automated positional systems. A club here may have two people and one camera. Applying major-league standards to a small collection system is an efficient way to manufacture fake data — because it forces the recorder to infer in order to fill out the form. I did exactly that during my first two years in the profession, and I call that the period when I wrote my best and was wrong the most.

The way I chose to handle it is not to reject weak data. It is to label it. In every report I now send, there is a column called "source confidence level," with three tiers: directly measured, inferred from video, and controlled guess. Only the first two are permitted in conclusions. The third may appear only in the assumptions section, and always accompanied by an open question.

The signal for the next cycle

I may be wrong about many things in this piece. My biggest assumption is that domestic basketball data will improve faster than the pace at which leagues expand, and that assumption has no evidence behind it yet.

But there is one signal I track and consider noteworthy: the share of reports containing at least one "insufficient data" column is rising among some young analytics groups. The people who write those words are not lazy. They are the ones who opened the raw file, saw 217 empty cells, and chose not to fill them.

If that trend continues, what changes will not be data quality — it will be the quality of the questions we dare to ask. And perhaps the most valuable question is not "how good is this player," but "who measured this player, when, and what were they told to leave out."