When the Numbers Are Blank: The Fragile Line Between Analysis and Fabrication

In a recent data meeting with the coaching staff of a professional basketball team, I presented an analysis table with twelve metric cells. Seven had numbers. Five were blank. The young assistant looked at the white space and asked what I would fill in. I said I would leave it as it was. The room went quiet for a few seconds. In sports analytics, a blank cell is usually read as a flaw, a sign of sloppy preparation. But years in this work have taught me the opposite: an honestly acknowledged gap is worth more than a number invented to fill the table. That is the first principle I hold, and also the hardest to keep in an industry where the pressure to look certain always outweighs the need to be right.

Modern basketball runs on data. Every game in the American professional league generates hundreds of thousands of tracking points, as camera systems log the position of the ball and ten players twenty-five times per second. European leagues, several of which I follow to cross-check methodologies, have caught up on collection infrastructure. Yet that abundance creates a subtle trap. When data floods in, analysts are tempted to fill every cell, including those the underlying sources were never reliable enough to answer. I once sat in a meeting where three different metrics on the same player were blended together simply because they looked related. No one checked how they were collected, by whom, or over what window. The table was full. The meaning was empty.

To understand why this matters, look at a specific case. In 2026, when I was twenty-five and working as an analysis assistant for a sports outlet in Hai Phong, I tracked a group-stage match between Switzerland and Serbia at a major tournament. I noticed that midfielder Granit Xhaka touched the ball one hundred and twelve times but played only thirty-four percent of his passes forward. I wrote a piece criticising his overly safe play, calling it a tactical regression. Coach Vladimir Petkovic pushed back, saying football is not mathematics. Three days later, Switzerland came from behind to win two-one on eight decisive passes at the key moments. I was wrong, and wrong in a systematic way. I had looked only at touches and forward-pass rate, ignoring PPDA, the measure of pressing intensity on the ball carrier, where Serbia ranked near the bottom of the tournament. When opponents press weakly, holding the ball and passing safely is optimal, not cowardly. A number detached from context led me to a conclusion exactly opposite to reality.

From that failure I set a discipline for myself: never conclude from a single metric. Whenever I assess a player, I force myself to check at least five baseline metrics, including PPDA, xG chain, pass progression, contested-shot efficiency, and chance-conversion rate. If one of those is unavailable or unreliable, I leave the corresponding cell blank instead of interpolating. A conclusion built on missing data will always be weaker than one that admits its own limits. This sounds simple, but it collides head-on with how sports media operates.

Look at the newsroom pressure. During a major tournament cycle, readers do not wait for caution. They wait for answers. Every time a big team prepares for a match, hundreds of analyses are published, and the most-read ones tend to be those that make decisive predictions. Decisiveness sells. Caution does not. A writer who says Team A has a sixty percent chance to win, with a wide confidence interval because two key players' fitness data is missing, is called timid. A writer who says Team A will win draws thousands of shares. This incentive structure pushes analysts toward manufactured certainty.

I was a victim of that structure myself. In November 2026, when I was thirty, a major newspaper invited me to write a prediction column before Saudi Arabia faced Argentina at a major tournament. Based on a model built from four years of qualifying data, I declared Argentina would win with ninety-four percent probability and a minimum score of three-nil. The result: Saudi Arabia won two-one, using an offside trap ten times in the first half that caught Argentina's attack offside seven times. My model had missed the variable no metric records: thirty-four degree Celsius heat and high humidity stretching the thigh muscles of South American players used to lower altitude and cooler air. After that failure, I spent two weeks rewatching forty-seven matches in Gulf tournaments over ten years to find the pattern. The lesson was not that the model was wrong. The lesson was that I had presented a number as a statement, when its nature was only a conditional estimate.

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

If sports analytics admitted this, the way we write would change from the root. Instead of starting with which team will win, we start with what we know and what we do not. Instead of ending with a verdict, we end with a quantified range of uncertainty. This is exactly where many in the industry object. They argue uncertainty makes analysis less compelling. I disagree. What makes analysis less compelling is hollow certainty, the kind that reality refutes days after publication. Readers can forgive a wrong prediction presented honestly; they do not forgive a wrong prediction presented as truth.

In basketball, the problem is more complex because many metrics are designed to look more precise than they really are. Take raw plus-minus. A bench player who enters for seven minutes while his team outscores opponents by ten will show a high plus-minus, yet that says almost nothing about his individual ability. It reflects teammates playing well, or opponents slumping. Time-weighted plus-minus was created to correct this, but it demands large samples and assumptions about lineup stability. When the sample is small, that metric can swing so wildly it becomes meaningless. Analysts know this. Readers do not, and the table does not announce its own limits.

This is where the idea of the number-chooser matters. Every metric has a person or group deciding what it measures, what it excludes, and how it weights factors. When I look at a performance ranking, my first question is not who is on top, but who designed the ranking and what story they wanted it to tell. Numbers do not lie, but the people who choose them do. A metric designed to reward fast-paced offense will automatically undervalue a disciplined defensive team. A metric favouring three-point efficiency will ignore the value of a centre who rebounds well. No metric is perfectly neutral, because every metric is the product of a choice.

In the 2026 season, when football and basketball paused for the pandemic, two colleagues and I built a set we called the Empty Court Index. The starting point was a question nobody could answer at the time: when no fans sit in the stands, how do players perform differently? We gathered data from two hundred matches in the Portuguese and Danish leagues after play resumed, then compared them with the same teams before the pandemic. The initial results were controversial: central midfielders' running distance fell nine point seven percent in the first month, but line-breaking passes rose thirteen point two percent. The club leadership I worked with was sceptical. They argued the sample was too small and the context too unusual to draw conclusions. They were right to a degree, and I told them so.

Still, I defended part of the analysis. If running distance fell while line-breaking passes rose, players might be saving physical energy for decisive bursts instead of running evenly all match. Based on that hypothesis, we recommended signing a Brazilian midfielder with a low-running profile but high acceleration. After ten rounds, he scored four goals and assisted three, including one from a fast counter-attack the model had predicted. The club climbed six places in the table. This story is often told as a data victory. But the more honest version is this: we were right because of a carefully built hypothesis on a small sample, and we knew clearly the sample was small. We did not hide the uncertainty; we consciously bet on it.

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

The problem for Vietnamese sports analytics lies here: we learned quickly how to use Western metrics, but we have not learned the culture of acknowledging their data limits that comes with them. Advanced stat tables are imported as-is, but collection conditions in domestic leagues differ substantially from the American or European game. The number of matches, the quality of tracking cameras, the way contested plays are logged, all affect metric reliability. A single expected-goals figure computed from fifteen matches in a semi-pro league can be so skewed it carries no statistical meaning, even if the formula is identical to the one used in a top European league.

I once witnessed such an error. An analysis outlet published a piece concluding that a young striker had superior scoring efficiency to every player of his age, based on an expected-goals metric. The number looked impressive. But when I checked its origin, it was computed from a sample of only eight matches, three against the weakest teams in the league. Remove just those three, and the player's ranking dropped to mid-table. The writer never told readers how small the sample was. That is not direct lying, but it is deliberate concealment of uncertainty.

From those experiences, I drew up a mandatory checklist before publishing any analysis. Where does the data come from. Who collected it. What was the method. What time window. How large is the sample. How many blank cells have I not filled. And finally, what could make my conclusion wrong. If I cannot answer those fully, I keep the piece in the drawer. This is a harsh discipline, and I admit I do not always keep it perfectly. But every time I break it, I pay with another instance of reality refuting me.

When the court is empty, only data whispers the truth. That is the line I write on the whiteboard in my small office in Hai Phong. It reminds me that most of a basketball game's story happens where cameras do not point: where the defenders move before the pass, who leaves the court last, which number never appears on the scoreboard. That data sometimes does not exist, or exists but is recorded by no one. And when it does not exist, the most honest choice is to acknowledge the gap, not to fill it with guesswork.

This is the difference between an analyst and a storyteller. A storyteller is free to tell a seamless story, because the goal is emotion. An analyst is bound by facts, even when the facts are a broken chain. The strength of analysis lies not in telling a complete story, but in daring to leave gaps that cannot be filled. An honest analysis is like a forensic report: it presents evidence, points out blind spots, and stops just in time before issuing a verdict the evidence does not permit.

There is a common objection: if analysts keep saying we do not yet know, readers will lose trust and leave. I think the opposite is true. Trust does not come from always being right, but from being honest about the possibility of being wrong. Throughout my career, my most loyal readers have been those who saw me admit mistakes. They return because they know that when I reach a conclusion, I have weighed its limits. Caution does not reduce authority; it builds authority more slowly but more durably.

In Vietnam's sports media market today, competitive pressure pushes many toward a faster path: making decisive predictions, writing sensational headlines, filling every cell with whatever numbers are available. That path brings short-term clicks. But it also creates a culture where data becomes decoration rather than a tool for seeking truth. And as readers slowly realise most decisive predictions were wrong, they lose faith in the whole field, including the careful work. The damage is collective, not individual.

The only way to break this cycle is to rebuild the relationship between writer and reader on methodological transparency. Publish the data source. State the sample size. Name the variables left out. Admit the confidence interval. These are unglamorous acts that produce no catchy headlines, yet they lay the foundation for a mature field. A mature field is one where people are unafraid to say they do not yet know, because they understand that acknowledging the unknown is the first step to truly knowing.

Back to that data meeting with five blank cells. After I explained why I would leave them, the head coach nodded and said something I have carried ever since: an honest table is more useful than a pretty one. We then spent two sessions redesigning the data system, adding tracking-based measurements, and turning blank cells into concrete tasks to be filled with evidence. Three months later, the table had no blanks. But this time, every number had a clear origin and every limit was noted beside it.

That is how I think about the future of Vietnamese sports analytics. We are at a stage where the tools are strong enough, but the discipline is not tight enough. We can compute the most complex metrics the world's top leagues use, yet we have not built the habit of questioning those metrics themselves. The distance between these two things is the distance between an industry that copies and one that creates. And the interesting part is that the leap does not come from a new algorithm or a larger dataset. It comes from a small cultural change: daring to leave a cell blank when there is not yet enough evidence to fill it.

With a major tournament cycle approaching, teams and analysts will again face the same familiar pressure: give answers before there is enough information. I will again sit before screens full of tables with gaps, and I know I will again be tempted to fill them. But I also know the greatest value I can offer readers is not a decisive prediction, but a framework rigorous enough for them to judge the reliability of every number they meet. Because in sport, as in every other field, the most important question is not what the result will be, but what we know, and what we still lack to know for certain.