Verify Before You Conclude: Notes from a Basketball Watcher

Late at night in Chengdu, I reopened the tracking file I had spent three weeks building. Sixteen games, over two thousand rows, every figure written by hand after each rewind of the footage until my eyes ached. When I scrolled to the final column, I froze: that column was empty. I had forgotten it while I was busy counting the things that were easy to count. That moment taught me more than a month of writing before it, and it is why I am writing this piece.

I tell that story because it repeats at a far larger scale. An entire sports-analysis field operates on tables that look full but are hollow exactly where it matters most. A reader sees a clean number, trusts it, and draws a conclusion. Very few stop to ask how that number was obtained, from where, and by whom.

Basketball analysis does not lack data. What it lacks is the discipline of verification.

There is a reason I keep returning to games nobody mentions. There, the data is cleaner, because it is not warped by crowd expectation and not haunted by big names. A game watched by millions generates millions of ways of telling it, and each telling lays another layer of interpretation over the original number. A forgotten game does not. That forgotten match taught me: football always speaks, only few bother to listen.

Ten years ago, when I was an analytics editor for a young sports site in Chengdu, a match the media overlooked gave me my first lesson. It was a fixture between Sichuan Jiuniu and Zhejiang Yiteng in China League One in 2026. Nobody replayed that game. Yet in it, a young defender, Huang Jiawei, shirt number 23, attempted 34 long forward passes and completed 27 of them, a 78 percent rate, against a league average of just 61 percent.

I wrote a piece on his role as a modern sweeper. Being a perfectionist, I revised it for a week. When it published, it caught the eye of a scout at a top-flight club, who later invited me onto the broadcast expert panel for the 2026 World Cup.

The lesson was not that opportunity. It was this: had I drawn a conclusion before re-checking the footage, the 78 percent would have been nothing but a pretty number. Clean data is not abundant data; it is data that has been verified.

From then on I built a three-step routine: cross-check the footage, cross-check the numbers, and cross-interview. No step is skipped. I build a personal data table before writing, focused on metrics the media rarely notices, such as completed long forward passes, screens set, or defensive redirections, instead of the numbers everyone repeats.

The cross-interview is the step I value most and the one fewest people take. When a scout and a coach say two different things about the same player, I do not pick a side. I record both, then look for why two people can read the same data and reach two conclusions.

People tend to think analysis is the work of numbers. Not quite. Analysis is the work of checking which number is trustworthy. A metric sitting inside a giant table can look highly professional, but if its source is vague, it is no different from a rumor set in bold.

In basketball this distinction is even sharper. Since the top North American league installed player-tracking cameras, the volume of data generated each night has grown beyond what any single person can read. That creates a temptation: to quote instead of to verify. Someone takes a number from an unclear source, places it beside another number of equally unclear origin, and calls it analysis. I have seen it, and I understand why it is dangerous: a wrong number is not merely a wrong number; it is a wrong conclusion waiting to be written.

Watching basketball and reading basketball are two different things. Many people only reread a game through the box score, then believe they understood it. The box score tells you who scored how much, but not the situation in which the points came, against whom, and at what moment. It took me years to understand that a player who scores twenty points has not necessarily played well, and a player who scores six has not necessarily played badly. The truth usually lies in the part that no column records.

Based on my experience following games, the most reliable thing is not the number on the screen, but the number I count myself and then encounter a second time in an independent source. Only when two independent sources agree do I allow myself to write.

I saw this more clearly in 2026, in the World Cup semifinal between France and Belgium at Krestovsky Stadium in Saint Petersburg. In the first half, I mispronounced the name of defender Toby Alderweireld three times. Viewers reacted online, but I did not argue. I spent a full month after the tournament reviewing footage of the 736 players at the event, building a standard transliteration list for every name, and analyzing how France's high press made Belgium's midfield triangle harmless. A three-thousand-word piece came out of it, later used as reference material by many young coaches at home.

Three mispronunciations taught me that a name matters less than the person behind it. With data it is the opposite: a name can be wrong harmlessly, but a number cannot. A single blank column can collapse an entire model, and I learned to check every cell before believing anything.

In 2026, when global football froze under the pandemic, I returned to Chengdu to work remotely. Sichuan Jiuniu, the club I had followed, fell into financial crisis, losing seven key players in one transfer window, including a striker who had scored 15 goals the previous season. While colleagues wrote emotional pieces about the club's tragedy, I quietly gathered liquidity data on sixteen League One clubs and compared it with the financial models of European second-tier teams.

I predicted Sichuan Jiuniu would finish eighth in 2026 and win promotion in 2026 if it held its academy together. Two years later, the prediction held down to the number.

The pandemic did not kill the club; the lack of vision killed it. That line reached me not from a speech, but from the rows of liquidity data I typed by hand over many nights.

Here is what I want to say against the grain. When a dataset is empty or contradictory, the usual response of an analyst is to pile on auxiliary hypotheses to save the model. I used to do that. I would add a variable, adjust a definition, move a boundary, just to keep the model running. That is exactly when analysis turns into fortune-telling dressed up in jargon.

I learned that an empty dataset is a gift, not a disaster. It forces me to disclose the broken part first, to admit the model is incomplete, rather than to hide it behind smooth prose. If a blank column in my own table could slip past me, then a hole in public data could slip past an entire industry.

The larger problem is that sports data now flows directly into places that should not have it. Live data fed to betting companies is the darkest side effect of the digitization of sport. A number released without a source, a chart published without a method, can become the basis for decisions its own creator never anticipated.

I still keep the old habit: build the table first, write second. Whenever I open a file and see a gap, I do not rush to fill it. I let it speak first. My position sits between the pitch and the truth, where not everyone dares to stand. And every deep analysis begins with a detail others overlook. The question for the reader should be: how did you verify that number.