The Night the Data Came Back Empty: Basketball Analytics and the Trap of Belief

Ten at night in Hai Phong, I opened the data file I had prepared for the next morning's tactics meeting. The file opened. There was nothing. The stat columns were blank, the player list was blank, the notes section was blank. An empty spreadsheet, in the literal sense. In nearly twenty years of analytics work, I had grown used to late data, wrong data, and data trimmed to fit a conclusion someone already held. But a completely empty file was new. And what chilled me was not the emptiness itself, but my own first reflex: to fill the gaps with memory.

I nearly did it. In my head the numbers appeared very clearly — how many points this player scored, who assisted the most, which team shot threes better. All of it plausible. All of it possibly true. And all of it backed by no evidence whatsoever. If I had not stopped, the next morning's meeting would have been built on a foundation that did not exist, and no one in the room would have known.

When data falls silent, the analyst must learn to fall silent with it — or to state plainly that he is speaking from memory, not from numbers.

That is the lesson modern sports analytics, especially in basketball, has to relearn every season. For more than a decade, the NBA has transformed from a league where a scout's gut decided almost everything into one where every decision — tactics, trades, even a star's rest minutes — must pass through a chain of data. Analytics departments have sprouted everywhere. Advanced metrics such as true shooting, estimated plus-minus, and win-probability models have become the shared language of coaching staffs and front offices alike.

But there is a paradox few are willing to voice: the more the industry depends on data, the more it can collapse because of data. Data does not create itself. It is collected by people, by cameras, by software, by processes that can break at any link in the chain. An empty file like that night's is not rare. It is merely rarely admitted.

I remember 2026, when I was an assistant analyst for a young sports outlet in Hai Phong. In the World Cup group match between Switzerland and Serbia, I eagerly drew a conclusion from a single number: Granit Xhaka touched the ball 112 times but played only 34 percent of his passes forward. I wrote that such play was too safe. Coach Petković answered curtly: football is not mathematics. Three days later Switzerland came back to win 2-1, with Xherdan Shaqiri scoring the decider, and I realized I had ignored PPDA — the pressure applied to the ball carrier — where Serbia ranked near the bottom. I was right about the number. I was wrong about the story.

That story taught me something that applies to basketball too: a metric has value only when you know the circumstances in which it was produced, by whom, and to answer what question. Basketball is the sport where data is distorted more easily than in any other, because pace creates countless different contexts for the same number.

Take the concept of "empty stats." A player averaging 25 points a night looks impressive. But if those 25 points come from a team losing game after game, in matches the opponent has already given up, on low shooting efficiency and a high turnover count, then the number does not describe ability — it describes opportunity. Opportunity to shoot, to hold the ball, to fail freely. Those are two entirely different things, and the box score does not distinguish them by default.

I always check at least five baseline metrics before believing anything: usage rate, true shooting, plus-minus, team pace, and opponent quality. Usage rate alone tells a different story. A player scoring 25 points on 30 percent usage is working under entirely different conditions from one scoring 20 on 18 percent. The box score will not tell you that. But if you never ask, you will forever believe the first is better.

Then there is garbage time — the closing minutes when the result is settled. This is where data is systematically inflated. A bench player scoring 12 points in the final 6 minutes of a 20-point loss looks like a discovery. But he scores against a backup defense, in a game state where no one is pressing, no one is committing tactical fouls. Feeding that number into a scouting model without adjustment is self-deception. Numbers do not lie, but the people who choose numbers do — and the person choosing here is the one who decides to record garbage time as if it meant something.

There is a deeper layer the box score almost never touches. When the game ends, the cameras switch off and the crowd goes home, and the truth lies in what no one recorded: when a defender rotated, who set the screen that created the gap, who made the cut the ball never reached. These are the numbers that never appear on the scoreboard, and precisely for that reason they are often the truest part of the game. Every number is a confession, if we are patient enough to listen — but the most important confessions are often never written down.

In 2026, when global football paused for the pandemic, I and a team of three built the "Empty Stadium Index" from 200 matches in Portugal and Denmark after the restart. We measured that central midfielders' running distance fell 9.7 percent in the first month, while through-balls rose 13.2 percent. The board was skeptical. I still convinced them to sign a Brazilian midfielder based on that model. After ten rounds he had scored four goals and assisted three, including a counterattack the model had predicted precisely. The team climbed six places in the table. New metrics are not born in offices; they are born in crises. When the world is upended, old measures lose their value, and we are forced to look at what no one bothered to record before. Basketball is the same: the best defensive metrics of today were born from the very games the box score could not explain.

But here is where I want to be counterintuitive. People usually believe the problem lies in wrong numbers. The more uncomfortable truth is this: the problem usually lies in correct numbers used wrongly. A perfectly accurate metric can still lead to a wrong conclusion if we mistake correlation for causation.

I once thought I was right. Qatar taught me I was wrong. In November 2026, before the Saudi Arabia versus Argentina match, my model combined four years of qualifying data and declared Argentina winners with 94 percent probability, by a minimum score of 3-0. The result: Saudi Arabia won 2-1, thanks to an offside trap sprung ten times in the first half. I had missed the most important variable — 34-degree heat and air pressure, which stretched the thigh muscles of South American players accustomed to playing at low altitude. I spent the next two weeks rewatching 47 matches in the Gulf region over ten years. My model was not wrong about the math. It was wrong about the world.

In basketball, that trap appears in subtler form. A team shoots threes well and wins a lot — we rush to conclude that three-point shooting is the path to victory. But it may be that defense and pace control are the cause, while the three-point rate is merely a consequence of leading. Correlation here is not causation. And the most dangerous part is that both numbers are correct. Only the story we tell about them is wrong. Data is a mirror; do not be angry when it reflects an ugly truth — be angry when we paint onto the mirror an image that does not exist.

That night, I did not fill the gaps. I closed the file, called the person in charge of data, and said the meeting had to be postponed. The next morning we had full data. Nothing exciting, no great discovery. Just the truth, as it always was.

In a season when every team is racing for data, perhaps the most valuable skill is not reading more metrics, but knowing when to stop and admit: I have nothing in hand. When the court is empty, only data whispers the truth. But when even the data is empty, the only truth left is the honesty of the analyst. The question for the next round is not which team has more numbers, but which team dares to look straight into the blank cells.