The Empty Report and the Analyst Who Refuses to Invent Numbers

On the night of July 14, 2026, in a small office on the seventh floor of a building in downtown Los Angeles, I opened a file named stage2_input.json. It contained nine sections, neatly divided: tactical and technical analysis, player data, team operations and salary cap, league landscape, rules and governance, coaching staff and locker room, risk, media and expectations, and the basketball industry ecosystem. Every frame was complete. Every cell existed. And the word inside every cell was the same: N/A.

The twenty-two-year-old intern sitting beside me, a master's student in sports analytics, finished reading and looked up with the face of someone who had opened a gift box and found air inside. "Do you want me to fill in estimates? I've read the news. I can guess."

I said no.

Two seconds to answer. Seven years in the profession to dare to answer.

Anyone who has sat in an arena on an ordinary night knows the feeling when the stat screens along the sideline suddenly go white midway through the third quarter. Nobody boos. Nobody leaves. The crowd just sits and waits, because everyone understands that a blank screen is not a blank game. The ball is still rolling. Only the data has gone quiet.

The empty report that night was the blank screen of my profession. And the greatest temptation for anyone in this line of work has never been exaggeration. It is filling the gaps.

An empty dataset is a signal, not an error to be concealed. The moment an analyst decides to fill a blank cell with an estimate is the moment he stops practising analysis and starts practising performance.

The summer of claims without foundations

July in America is Summer League month. In Las Vegas and San Francisco, teams put unproven players on the floor for ten days, and in those ten days a small industry is born: the industry of conclusions.

Every night there are six games. Every game features roughly twenty players in jerseys, half of whom will never touch an official NBA game. But within twelve hours of the final whistle you will read hundreds of decisive assessments: this player is a "hidden gem," that one is a "scouting failure," this team "won the offseason," that one "collapsed."

I am not against making judgments. I live on judgments. But I have tracked enough Summer Leagues to notice an uncomfortable pattern: the quality of a July conclusion is usually inversely proportional to the length of the article containing it.

Shorter pieces, harder claims. Harder claims, thinner evidence.

Beneath that flood of assertions lies a volume of data unprecedented in basketball history. The NBA has captured player-tracking data since the 2026-14 season, when SportVU was installed league-wide. From 2026-24, Hawk-Eye replaced it and pushed capture frequency beyond what the human eye can follow. An ordinary game now generates positional, velocity, trajectory and distance data points that no coach could read in a lifetime.

The paradox sits right there. More data than ever, yet the reliability of popular conclusions has not risen with it. It has moved the other way. When the cost of producing a claim approaches zero, the market will produce claims until there is no room left.

And when every space is filled, empty space becomes the most suspicious thing of all. Nobody wants to hand in a blank sheet.

Anatomy of an empty data layer

The report from July 14 was not the product of laziness. It was the product of a failure one layer upstream, and it accidentally became the most honest document I have read this year.

In the system I run, every analysis passes through two layers. The first layer decomposes source text into information points: events, numbers, subjects, timestamps, provenance. The second layer takes those points and builds nine dimensions of deep analysis. This is the architecture I learned after my first career shock, and it performs well in most cases.

But every architecture has a point of death.

When the first layer returns an empty set — no title, no source, no core viewpoint, no recognised entity — the second layer has exactly two choices. Stop and report the failure. Or fabricate.

I have watched both choices made in this industry, and I know the price of the second.

Four causes produce an empty data layer, and they demand four entirely different responses.

The first is a genuinely empty source. The original article sits behind a paywall. Or it is video, a podcast, a format the machine cannot read. Here what is missing is access, not data quality.

The second is extraction failure. The text exists, dense with information, but the extractor fails because of an unusual format, an unsupported language, or a small structural change in the document. This case is more dangerous, because it produces an empty set that looks identical to an article that truly has no content.

The third is a contract failure between layers. The first layer returns the right fields, but the second reads the wrong names, or the first changes its structure without notice. This is a purely systemic fault, and it is the most expensive kind, because it is silent and recurring.

The fourth is time pressure. The deadline arrives before the data does. And this is the only one of the four that is not a technical fault. It is a human one. It is why I told my intern no.

These four causes lead to four different actions: request access, fix the extractor, patch the data contract, or move the deadline. None of them is filling a blank with an estimate.

When a system returns "insufficient information," it is telling you something more precise than any number: it is telling you that you have no basis for a conclusion. That is information. Not a gap.

Three stories that taught me to be silent on time

My career was built on three failures, and all three belong to the same family: I had correct data but did not know how to make it speak at the right moment.

In 2026, I was twenty-four, newly hired at a basketball analytics blog in Los Angeles. At that year's Summer League I found that Dillon Brooks — an undrafted free agent — posted an individual defensive rating of 98.3 across five games. The man competing for his roster spot, Troy Williams, managed only 104.2. Six points in a defensive metric is a gap wide enough to say the two men defend at different levels.

I knew I was right. I also knew I was not certain enough. So I spent three weeks refining a probability model, cross-validating, tuning weights, rerunning every scenario. Three weeks.

A rival blog published a piece celebrating Dillon Brooks three days before me.

Nobody read mine.

The first lesson was not "write faster." The lesson was to redefine "good enough." Since then, every analysis I produce has a finished draft forty-eight hours before deadline, with the final twenty-four hours reserved purely for verifying numbers. I no longer chase infinite perfection, because I have seen it pay wages in silence.

The second story arrived in 2026. When the NBA shut down for COVID-19, I spent four months researching the history of hamstring injuries after long layoffs. I found a pattern: after an extended competitive break, the risk of hamstring re-injury rose by roughly 1.6 times if a player returned on a two-games-a-week schedule.

Kawhi Leonard sat in the highest-risk group.

I wrote a forty-page report and sent it to the LA Clippers medical staff. Forty pages. Charts. Appendices. Models. Everything correct, thorough, dense.

Nobody read it through.

In August 2026, Kawhi Leonard suffered exactly the injury the model predicted, and the Clippers exited the playoffs in the second round.

Nobody read the report on Kawhi's knee. The market only read it after the sound of the snap.

The second lesson was not "don't write long." It was the art of the summary. Every document I have written since opens with a one-page executive summary, the recommendation on the first line, with the full evidence behind it for anyone who wants to check. A sporting director must finish it in two minutes and know exactly what to do. If he has to reach page seventeen to understand my point, that is my failure, not his.

The third story came in 2026, at the World Cup in Qatar. A brokerage asked me to assess South American talent. I applied the early-signal framework I had refined four years earlier and identified Enzo Fernández, then at Benfica, on two standout metrics: 11.4 progressive passes per ninety minutes, and a 78 percent success rate under pressure — the best among under-23 midfielders at the tournament.

I wrote a two-page report. Two pages. A recommendation to sign him for thirty million euros.

In January 2026, Chelsea bought Enzo Fernández for one hundred and twenty million euros. My report leaked onto a data forum.

Three stories, three different failures, one shared lesson: the value of an analysis does not lie in its accuracy. It lies in reaching the right person, at the right time, in the right format.

After the 2026 leak, I set an internal rule: every internal report codes player names as numbers. Only when a contract is signed does a real name appear in a document. Not because I fear losing ownership of analysis. Because a leaked name can change the transfer value of a twenty-two-year-old human being.

More data means errors spread faster

Here I must say what most of the industry does not want to hear.

Over the past decade, basketball analytics solved the problem of collecting data and failed at the problem of communicating it.

We have positional data for every player in every hundredth of a second. We have shooting efficiency metrics, estimated impact metrics, opponent-adjusted defensive metrics. But most of those numbers reach the public through a three-layer chain, and each layer strips away part of the truth.

The first layer is the data room, where a number is born with full context: sample size, margin of error, collection conditions, confidence level.

The second layer is the press room, where the number is severed from context and becomes a headline.

The third layer is social media, where the headline is severed from its source and becomes a belief.

By the time a number travels all three layers, it is usually stronger than the truth and thinner than the context.

This explains something I have tracked across many seasons: fans know more metrics than ever while evaluating players less accurately than ever. Knowing a metric is not the same as understanding it. And understanding it is not the same as knowing when to discard it.

Data is like a book. The crowd looks at the cover. The wise read every page.

Against that backdrop, an analysis layer returning an empty set has surprisingly high diagnostic value. It forces the operator to stop. It does not let the three-layer chain continue flowing. It blocks a false conclusion before that conclusion can become the belief of a hundred thousand people.

I checked that empty report three times on the night of July 14. No fault in the second layer. No wrong parameters. No broken model. The cause sat in the first layer, and it belonged to the first of the four categories: an unreadable source. The original document was a video news item that had never been transcribed.

The only correct conclusion available was: no conclusion yet.

So I wrote that down.

The other side of confidence

There is a professional pressure nobody talks about, and it costs more than any consulting fee.

The sports market pays for decisiveness. Nobody pays for the sentence "I do not have enough data to conclude." In a boardroom, the person who says "this team will make the playoffs" gets a note taken. The person who says "playoff probability sits between 48 and 62 percent, and that range is too wide to act on now" is read as indecisive.

This asymmetry has concrete consequences. It pushes the best analysts toward two escapes: overclaiming, or going fully silent.

Both escapes cause damage. The first produces conclusions that cannot be trusted. The second produces rooms where data has no voice.

I chose the second escape for years. I had data, I had models, but I did not publish because I feared being insufficiently certain. The result was that my findings sat in a drawer until reality confirmed them, and by then they were useless.

My article was late not because I was wrong, but because I did not believe in myself enough.

This asymmetry has a more dangerous variant, and it connects directly to the most ignored subject in every NBA injury debate.

When a star goes down, the market immediately hunts for a single cause: one collision, one wrong movement, one bad landing. But the data I have collected across many seasons points elsewhere.

Schedule density is the biggest culprit. No medical staff can save a player who plays two games a week for six months, crosses four time zones, and accumulates more than ten thousand miles of travel per month. A bad landing is only the final moment of an overload chain that has run for twenty games.

Schedule density belongs to the category of information nobody wants to hear. It has no imagery. It has no slow-motion replay. It cannot be cut into a fifteen-second clip. It is just a calendar.

And a calendar never makes the news.

That is why I spend most of my professional time on health and workload reports that broadcast departments call "dry." Those reports have never gone viral. But they are right. And being right does not require going viral to become true.

Data that is correct but ignored is not data — it is a debt owed by those who refused to read it.

When a finding becomes a fact

There is a question I always ask myself before publishing anything, and I think it should be the industry standard.

If I am right, on what date will that be proven?

Not "am I right." But "when, and by what data."

A judgment without a verification checkpoint is not a judgment. It is a feeling wrapped in the language of statistics. And a feeling disguised as statistics is the hardest kind of misinformation to detect, because it has the full outward form of a scientific conclusion without the interior.

Every finding needs a moment in time before it becomes a fact.

I apply this principle even to small things. When I say a young midfielder has strong progressive-passing numbers, I must specify how many games I expect that number to hold and the threshold below which I will admit I was wrong. When I say schedule density is pushing a star to his limit, I must specify the window in days and the signal that will confirm or refute it.

This principle has a side effect I learned rather late: it makes admitting error much easier. Once you have declared your falsification conditions in advance, being wrong is no longer a shameful event. It is a step in the process.

Conversely, someone who never sets a checkpoint will never admit error, and therefore never learn. They only need to edit their memory. And the market will help them do it, because the market has a very short memory.

The unknown as an asset

I want to return to the nine empty sections in the July 14 file, because I believe they deserve to be read as a professional document rather than an incident.

In tactical analysis, an empty set means no system was identified, no lineup recorded, no offensive or defensive efficiency metric supplied. An honest analyst writes exactly that.

In player analysis, an empty set means no subject to place on an age curve, no shooting efficiency, no usage rate, no contract data. An honest analyst assigns no position to a player who does not exist in the data.

In salary cap analysis, an empty set means no max contracts, no mid-level exceptions, no trade assets identified. An honest analyst builds no financial scenario out of nothing.

And so it continues across nine sections.

My point is this: a document that states ten times that it does not know has higher professional value than a document that states ten times that it knows, if the second is built on guesses.

The reason is practical. In a multi-layer process, a small error at the first layer is amplified at every subsequent layer. If layer one invents a number, layer two builds a model on that number, layer three issues a recommendation based on that model, and layer four carries the recommendation into a decision worth tens of millions of dollars.

Errors do not add. They multiply.

That is why I told my intern no. Not because I doubt his ability to guess. Because I believe in multiplication.

What I write today may be forgotten. But the system it builds will not be.

Variables for the coming weeks

Summer 2026 is still running. Teams are still finalising rosters, young players are still chasing a spot, and data rooms are still running forecasts for next season.

There are three checkpoints I will hold myself to over the next six weeks.

The first is the official rosters before the season opens. I have recorded my prediction that at least two undrafted players from this year's class will land on the roster of a playoff team. If that happens, I will be right on the conclusion. If not, I will be wrong on the conclusion, and I will rewrite that section rather than rewrite my memory.

The second is injury reports in the first two months of the season. I predict that hamstring and calf injuries among players averaging over thirty minutes per game will exceed the three-season average, provided the current schedule density holds. This is a claim falsifiable with public data, and I accept that.

The third is the quality of this very report. Over the next thirty days I will check how many "insufficient information" fields in my input files were resolved by fixing the process, and how many were resolved by filling in an estimate. If the second number is greater than zero, I have a system problem. If it is zero, I have a system.

I do not expect to be right on all three. I expect to know exactly whether I was right or wrong on the day I declared.

That is the entire difference between a judgment and a hunch.

In a summer when everyone has a conclusion, the scarcest thing on the market is not a better conclusion. The scarcest thing is a person willing to hand in a blank sheet, knowing exactly why it is blank, and knowing precisely the date they will return to fill it with a number that can be verified.

My intern will learn that. Not from tonight. But from the night he has to choose for himself between an honest empty cell and a beautiful number.