Full Spreadsheets, Zero Evidence: The Hollow Disease of Basketball Analysis

Last March, a friend working in the analytics department of a B.League club sent me a 41-page document. It was complete to an uncomfortable degree. A ball-advancement assessment table. Offensive and defensive efficiency metrics. A personnel-fit analysis box. A section titled 'playoff transferability.' And a risk matrix with full columns for probability, severity, impact, and mitigation. Every cell was colour-coded. Every table had a bold header. It looked like a finished product down to the last detail.

I read all 41 pages, closed the file, and realised there was not a single player's name in it.

No team name. No coach's name. No date. No cited source. Not one specific metric with a unit of measurement. The only fully populated field was the domain label: basketball.

That was the moment I understood I was holding something more dangerous than a wrong analysis. A wrong analysis can be argued with, corrected, learned from. This thing looked so right that nobody would bother checking it. And it repeats across the industry, far beyond one isolated case.

In basketball, the analytical structure was standardised long ago. An NBA scouting report typically has nine sections: tactics, player data, cap operations, league landscape, rules and governance, coaching staff, risk, media narrative, and industry ripple effects. The B.League copied it. Analytics centres across Asia copied it. The framework spread fast because it is convenient: open the template, fill it in, export it. The problem is that the framework still renders fully even when the input data is empty.

In 2026, I walked straight into this trap.

Before the Tokyo Olympics, I wrote a long piece predicting the Japanese men's national team would reach the quarterfinals. My reasoning was the arrival of Rui Hachimura and Yuta Watanabe, the country's first NBA players, alongside what was rated the best preparation season in Japanese basketball history. I staked my reputation on that prediction.

Japan lost all three group-stage games. The 77-97 defeat to Argentina was the one that made me sit down. Their defensive rating in that tournament was 118.4. I had seen that number before I wrote the piece. I simply did not put it at the top.

I wrote a 1,500-word public mea culpa. And from that day I set myself one rule: never make a prediction based on a player's reputation. Reputation is yesterday's story. Today's metrics are the truth.

But it took that 41-page document for me to see the other side of the problem.

The greatest danger in modern basketball analysis is not a lack of data — it is a complete framework with no data behind it. A table with headers, columns, colour, and footnotes triggers an automatic credibility response in the reader's brain. Psychologists call it the fluency effect: what is easy to read is easy to believe. A neatly formatted report is rated higher than a correct but messy one.

On a basketball floor, the damage is concrete. A coaching staff reads the report, sees a risk section highlighted in red with a severity rating, and concludes that risk has been assessed. Nobody checks what is actually inside that section. The deliverable has been shipped. Responsibility has been transferred. And nobody is accountable for an empty cell painted red.

In player evaluation, the concept of empty stats is familiar: a player averaging 22 points for a team that loses 55 games. Pretty numbers, low value, because context was ignored. In the NBA they call it good stats, bad team. What few people say out loud is that analysts produce empty stats too. A report dense with tables, heavy with jargon, carrying a risk matrix and bar charts — yet containing not one verifiable data point — is exactly that 22-point average on a 55-loss team. It looks like work was done. No work was done.

The only way to tell the difference is to count traceable data points, not filled cells. This is the test I apply to every report I receive, and to my own writing before I hit publish.

Every basketball analysis has a mandatory minimum threshold. There must be at least one named entity — a team, a player, a coach, a governing body. There must be at least one quantitative data point — a metric, a salary figure, a record, a percentage. There must be a defined time marker. Without those three, everything else is decoration.

That 41-page document had all nine sections. It lacked all three minimum conditions.

I call this an industry problem rather than an individual one for a specific reason: once a framework is standardised, people start measuring the quality of work by how complete the framework is. An analyst gets praised for submitting a report with every section filled. Nobody gets praised for submitting a report with three data points nobody else has. The incentives flow toward form, and form is always easier to produce than substance.

After the Tokyo shock of 2026, I built a three-pillar evaluation framework: offence, defence, and conditioning. Each pillar requires at least two quantitative metrics drawn from a minimum of five games. If the data is insufficient, I write 'insufficient sample.' Those two words have saved me more often than any spreadsheet I have ever built.

Applied to Japan's 2026 squad, that framework produces a very different conclusion from the one I published. Offence: solid scoring efficiency but dependent on two individuals, with no fallback system when one of them was face-guarded. Defence: a defensive rating of 118.4, well below the tournament average. Conditioning: a clear fourth-quarter decline across all three games. Three pillars, two flashing red. The correct conclusion should have been that the quarterfinals were highly unlikely. I did not reach it, because two NBA names had already filled the empty cells in my head before the data had a chance to speak.

There is a psychological mechanism more frightening than laziness: the pressure to fill blank cells. When you sit in front of a table with 60 cells and 58 already completed, the two empty ones exert an invisible pull. You want to fill them. You will fill them. And you will fill them with what sounds most plausible, not what is most accurate.

I once reviewed a scouting report that described a young guard's defensive metrics as adequate. When I asked for the source, the author said he had watched three games and it felt that way. Three games is far too few. In my own data, I will not make a claim about a player without at least five games for cross-reference. Three games can be a lucky streak. Five games start to reveal a trend. This is the ethical line of the profession. Data does not lie, but the people who read it do. And the people who write it do too.

When an empty analysis enters a system, it does not stop there. It reaches the coaching meeting. It is used to justify a rotation change. It causes a young player to be undervalued, or overvalued. Worst of all, it becomes precedent: next season the framework is reused, and emptiness becomes the standard. I remember an Asian club that paid 140,000 USD for an international data provider's package and received a 300-page report set. A year later they still had not changed their style of play, because the report could not identify a single player who needed replacing. The failure of a giant is a gift to the observer — but only if the observer bothers to read the footnotes.

Here I have to argue against myself a little, because it is easy to fall into the reverse trap: assuming every report full of tables is garbage. That is wrong. A standardised framework has genuine value. It forces the analyst through questions intuition tends to skip: what about the salary cap, what about the rules, what about risk. The problem lies in the order of operations, not in the existence of the framework.

On the other hand, an analysis with few tables is not automatically correct. I have read very short, very sharp pieces whose conclusions rested on a single game. Conciseness is not accuracy. Brief and empty is still empty — just empty faster.

My chosen approach: keep the framework, but reverse the process. Start with data points, then build the table. If after three data points a section still cannot be filled, that section should be left blank — and explicitly labelled blank due to insufficient data. An honest empty cell is worth more than a fabricated full one. The risk of this approach is being seen as doing less. I accept it. In nine years of covering this industry, I have never seen anyone fired for saying they did not have enough data. I have seen people fired for saying something with certainty and being wrong.

Great teams collapse not because they are weak, but because they forget they were once small — they forget they once had to count every metric by hand instead of buying a stack of pre-packaged reports. Empires are not built in a night, but data can build them in a single season. The point is real data, not handsome spreadsheets.

Next game you watch, try one thing: after reading any analysis, ask yourself how many names and how many specific data points you can recall. If the answer is none, that analysis never began. And if you are the one writing it, that question is far harder.