The N/A Trap: Modern Basketball Is Drawing Conclusions From Empty Data Cells

In the dashboard I open every night to follow the NBA, there is one column I have learned to fear: the empty one.

Zero is a measurement. It says that something did not happen during the observed period. An empty cell says something entirely different: that it was never measured at all. Over fifteen years sitting in front of basketball spreadsheets, I have watched those empty cells get read as "no problem here" hundreds of times. A player with no standout defensive metrics becomes a bad defender. A team that publishes no injury data becomes a healthy team. A game with no hustle numbers becomes a game nobody defended.

On February 18, 2026, in Indianapolis, the All-Star Game ended 211-186 for the East. Forty-eight minutes, 397 combined points, and almost the entire American basketball media reached the same conclusion: defense is dead in an exhibition. I do not dispute the conclusion. I dispute how it was produced. The tracking boards that night recorded very few contested shots, very few collisions, very few defensive turns, simply because nobody on either side was actually trying to defend. Data about effort did not exist. What existed was data about outcome: the ball went in. Then people used the absence of effort data to prove that effort had vanished.

That is the trap I want to write about. Not the trap of wrong numbers. The trap of empty numbers.

Context: Ten years of turning basketball into a maths problem

To understand why an empty cell is so dangerous, you have to remember how the sport got here.

In 2026, a handful of NBA teams installed STATS SportVU: six ceiling-mounted cameras recording the position of the ball and every player twenty-five times per second. By the 2026-14 season the whole league used it. For the first time in the sport's history, people knew exactly how far a player ran, at what angles he turned, and how far he stood from a teammate when he caught the ball. Things that had lived only in the viewer's eye became numbers you could print.

In 2026 the NBA selected Second Spectrum as its official player-tracking provider from the 2026-18 season, replacing the SportVU era. With it came a new generation of metrics: the probability a shot goes in under specific conditions, the expected value of a possession, the attention a shooter draws from a defence. From the 2026-24 season, the league deployed Sony's Hawk-Eye skeletal tracking to automate part of its statistics, starting with rebounds, assists, blocks, steals and deflections.

For Vietnamese fans, this revolution arrived later and through a narrower door: box scores, shot charts, and heat maps shared onward through social media. We almost never see the raw data, never see the notes about missing measurements, never see a question mark placed beside a suspicious figure. We see the tip of the iceberg, and the tip always looks clean, tidy, as though it has explained everything in the world.

I entered the profession exactly as that iceberg surfaced. In 2026 I was twenty-one, a final-year economics student in Miami, and my first professional lesson did not come from a spreadsheet.

Euro 2026 taught me something: a hot take does not need to be right, it needs to be timely. On the night of the final, in a bar, when Cristiano Ronaldo left the pitch injured in the twenty-fifth minute, I blurted out to the friend beside me that Portugal played better without Ronaldo and would win the trophy. The bar laughed. Portugal won 1-0 through Eder. I collected forty-seven dollars in winnings and a belief: going against the consensus is a skill you can train.

But that belief had a dark side it took me years to name. When you go against the consensus on a hunch, you are playing a probability game. When you go against the consensus using a dataset full of holes, you are playing a game of fooling yourself.

Heat maps have become a new kind of fortune-telling

Take the most familiar artefact: the shot heat map.

A heat map tells you where a player shoots from and how efficiently. It is a good measurement, geometrically precise. The problem is that people read it as a description of ability, when it is only a description of choices.

A player shooting forty-two percent from the right corner is not necessarily a great right-corner shooter. He may simply be the beneficiary of a system that always creates an open right corner, thanks to a teammate drawing two defenders at the top. The heat map does not record that teammate. It does not record that the defence chose to leave him. It does not record that he turned down four hard shots in the fourth quarter to wait for the easy one.

This is why I say heat maps have become a new kind of fortune-telling. They project certainty onto something that depends entirely on the system around it. And when the heat map is empty, when a player has taken only twelve shots all season, people immediately conclude he has no weapons. The truth may simply be that he was never given the ball in a good enough position to show any.

Nikola Jokic is the cleanest example of this trap, in reverse. For several seasons he sat among the slowest players in the league by average speed. That number is correct. He really does run very little. But it cannot measure the thing that makes him the centre of an entire offensive system: his ability to read the floor before the defence has moved. A speed field has nowhere to record that he just saw a pass nobody else on the floor had seen.

Read the speed table and you conclude Jokic is slow. Watch him play and you see that speed is not the tool he uses. Both statements are factually true, but only one of them explains the game.

The same thing happens in youth academies. A big club may sign twenty prospects a year, and the academy database will look dense, impressive, full of metrics. But if you track where those players actually go, most will never touch the first team. A talent-hoarding system produces two kinds of empty cells: the cell of a player never recorded because he never played, and the cell of a player over-recorded at youth level before vanishing as a professional. Neither cell shows up on the transfer ticker.

The system measures position, not intention

Here is the limitation at the centre of the entire analytics industry, and it is discussed far less than it deserves.

Tracking cameras record coordinates. They know where a player is, where he is going, how fast. They do not know what he is thinking, what he is waiting for, what he is trading away.

A simple example: two defenders each run 4.2 kilometres in a game. One is in the right place, closes the right gap, forces the offence to change direction and make a bad decision. The other chases the ball like a dog chasing a car, always half a step late, and his distance number is higher precisely because he has to make up ground for his own bad choices. The box score gives these two men nearly identical results. One is better. But the field that would record that superiority does not exist.

This is why individual defensive metrics in basketball are always more suspect than offensive ones. You cannot measure a successful defensive possession, because its success is usually that the opponent achieved nothing. Achieving nothing generates no data. It generates an empty cell, and that cell then gets filled with a conclusion the writer invents.

Since the NBA moved to Hawk-Eye automation for rebounds, assists, blocks and steals from the 2026-24 season, the precision of the measurement has improved. But the gap between measurement and meaning is untouched. An automatically logged deflection still does not tell you whether the player read the pass or merely stood in the right place by accident.

There is one concept teams use to patch this hole: gravity. When a shooter stands in the corner, two defenders may be forced to keep an eye on him rather than on the ball. Tracking data can measure the distance between defender and shooter and infer attention from it. But it is still inference, not observation. And inference depends on the model, and the model is written by people, with assumptions people chose.

When I worked for a Miami sports podcast in 2026, I kept a private board of entirely non-traditional metrics, refreshed every morning. The point was not to understand basketball better. The point was to fire off a take before the crowd saw the number. I predicted Damian Lillard would be king of the playoff run in a crowdless environment, based on twelve scrimmage games inside the bubble. It was right. Lillard carried Portland into the playoffs and had nights over fifty points. But looking back, I knew I was betting on a twelve-game sample, and I knew that if it failed I would lose nothing but an article.

That is the privilege of the writer. A team does not have that privilege.

Expected metrics are not measurements

There is a common confusion I meet almost weekly in Vietnamese basketball writing: treating expected metrics as objective measurement.

The expected value of a possession, the probability a shot goes in, any metric with the word expected in its name, is the output of a statistical model. That model is built from historical data and answers a very narrow question: under conditions like these, what percentage of the time did people score in the past.

It does not answer who is shooting, in which quarter, with which wrist.

When a model produces a number, that number is only as good as its inputs. If the inputs are missing, the model does not throw an error. It does not display an empty cell. It returns a number that looks entirely normal. That is the most dangerous feature of the whole system: a data gap at the root becomes a confident conclusion at the branch.

I have seen this happen to me. In 2026, working for a small Miami sports channel, I was sent out as a fan reporter to a public viewing area during the World Cup in Russia. During the semi-final between Croatia and England I mispronounced Luka Modric's name three times in a row on a livestream. Views dropped. But when Croatia came back to win 2-1, I immediately published an analysis arguing England lost because they feared the long ball rather than because they were tired, citing thirty-eight Croatian long passes against eleven for England. That piece was shared two thousand times in twenty-four hours.

World Cup 2026, I mispronounced Modric. That whole night I learned about the twist. And I learned something else: a number chosen at the right moment can hide the fact that I never checked whether that number actually explained the game.

That is why I believe the biggest problem in modern basketball analysis is not a shortage of data. It is the surplus of data generated from regions where we do not know that we measured nothing at all.

Small samples, the bubble, and forgotten honesty

NBA Bubble 2026 had no crowd. All I could do was listen to myself.

That is what I wrote when the league returned in Orlando. Every commentator was panicking because there was no arena noise to lean on, no atmosphere to describe, no crowd to blame. I dived into historical data and found something interesting: in crowdless environments shooting metrics became more stable, less volatile. The cause might be that isolation reduces psychological noise. The cause might also be that the sample was far too small to conclude anything.

I chose to write it as though it were the former. It worked as media. It was not honest as method.

The problem with small samples in basketball is not that they produce wrong numbers. The problem is that they produce correct numbers about something too small to matter, and readers have no way to tell those two things apart unless the writer says so. A player who makes four of five threes over the first games of a season is shooting eighty percent. That number is true. It is also meaningless.

But in sports media, saying a number is meaningless generates no shares. Saying a player is exploding does. This is the structural pressure that pushes analytics toward filling empty cells with better stories instead of leaving them empty.

Sports culture is an endless argument after the final whistle. And in that argument, the loudest voice usually has the least evidence.

Sixty-five games and the limits of health data

No area reveals the empty-cell trap more clearly than load management.

From the 2026-24 season, the collective bargaining agreement signed by the NBA and the players' union in 2026 introduced a rule requiring players to appear in at least sixty-five regular-season games to be eligible for major awards such as MVP, All-NBA and Defensive Player of the Year. The same year, the Board of Governors approved the Player Participation Policy, tightening rest rules around nationally televised games with fines starting at one hundred thousand dollars for a first violation and escalating after that.

Both rules rest on an implicit premise: that one can measure a player's fatigue and injury risk and optimise minutes accordingly.

In reality, health data in professional sport is far thinner than the public imagines. Teams track workload, acceleration, hard decelerations, jump landings. They have data on what the body has done. They have very little data on the body's internal state, on soft tissue, on sleep, on a player pushing through a small injury nobody has detected.

So there are regions that are entirely empty, and those regions get filled with the judgement of the medical staff, the pressure of the coaching staff, the player's own desire to play, and the sixty-five-game rule just written into law.

When Joel Embiid tore his meniscus in late January 2026, the immediate question was whether the sixty-five-game threshold had pushed him onto the floor before he was ready. I have no data to answer that question. And I notice almost nobody does. There is only an empty cell, and into that cell people place whatever they already wanted to believe.

This is where rules collide with data. A rule needs a measurement. The measurement does not exist. So the rule operates on a substitute measurement: games played. It is a real measurement, clean, indisputable, and almost entirely unrelated to the thing it claims to measure.

Where I might be wrong

I have to write this section before the conclusion, because otherwise this whole piece is just a long hot take wearing technical vocabulary as a costume.

Possibility one: the gap closes on its own. Every year tracking becomes more precise, cheaper, more camera-rich, more machine-learned. Perhaps in five years measuring intention in defence will be routine, and my scepticism will be filed under the nostalgia of a writer getting old. I accept that risk. The history of basketball analytics is a history of sceptics being proven wrong slightly later than the moment they spoke.

Possibility two, and this one bothers me more: the person producing the most empty cells may be me.

I make a living issuing judgements before the evidence is complete. I built a career on the principle that a timely take outlives a correct one delivered late. Over fifteen years I have many times used a factually correct number to prop up a conclusion that was wrong in substance. I have done the thing I am criticising here.

I forge hot takes, but the truth is the thing I have forged the longest. And the hammer is not always in my hand.

Possibility three: perhaps the empty cell is not the problem but the nature of every sport. Basketball has always contained things that cannot be measured. Before SportVU, people still told magnificent stories about defence, about locker-room leadership, about a bench player changing a team's personality. Perhaps the truth is that basketball is at its best when we do not know everything, and my demand for a hole-free analytics is a demand for something both impossible and dull.

I am not sure I can argue against that.

What I expect to happen

If you ask me what changes in the next three years, I will not say cameras or models. I will say a job title.

NBA teams will soon have to hire people whose only job is to check whether the data they are using was actually measured. A department that reconciles measurement against absence. And when that title appears, there will be a counter-intuitive consequence I want to see: team dashboards will look worse. More empty cells will be flagged rather than filled with estimates. Reports will get longer, slower, less exciting.

That is the sign of an industry maturing, not weakening.

Until then, I will keep one small habit when reading any piece of basketball analysis, including my own: find the empty cell before finding the number. Anyone can cite a metric. A decent professional is someone who can say which metric is missing, and how the conclusion changes without it.

I do not write to be right, I write to open a corner nobody has looked at. But opening a corner with a map full of holes just means the hole gets retold very loudly.