The Empty Analysis: When Table Tennis Refuses the Writer a Guess

The Empty Analysis: When Table Tennis Refuses the Writer a Guess

6:12 a.m. in Guangzhou, the city still damp with mist outside a tenth-floor window. The tea in the cup had gone cold long ago. On the screen sat a file just delivered by the two-stage analysis system — the kind of file I run every day to sweep the sports wires before I sit down to write.

The file has eleven fields. The article-title field is empty. The source field is empty. The one-sentence-summary field is empty. The list of information points — which should hold two to four concrete items — is completely blank, not one line. The list of entities involved — players, coaches, associations, events — is also blank, because it is designed to draw from the information points above it, and above it there is nothing to draw from. The time-sensitivity field states plainly: not assessed. The source-quality field states: not assessed.

Only one box was lit: the domain label — table_tennis.

A data file announcing that it belongs to table tennis, and then falling silent.

Thirty years in this trade taught me to separate two very different things: a data file that is thin, and a data file that is empty. A thin file still offers something to dissect, to doubt, to down-rate in line with its sample size. An empty file permits exactly one honest action: stop, and record that you stopped.

In this industry, that stopping is the hardest thing to write.

Two stages, nine dimensions, one empty object

The system I use runs in two stages. Stage one reads the raw article and decomposes it into a structured object: title, source, article type, one-sentence summary, author stance, purpose, information points, the list of people and organisations, time sensitivity, source quality. Stage two takes that object and applies nine professional dimensions used in the table tennis domain: technique, tactics and equipment; player data and head-to-head records; the event system and points rules; the competitive landscape between nations; rules and governance; coaching staff and the talent pipeline; the risk surface; public narrative and expectation; and industry transmission.

That morning, stage two still ran. It ran all nine dimensions, every table, every risk label, every confidence grade. But because stage one returned an empty object, every data cell in stage two carried the same line: insufficient information to assess.

The result was a long, structured, layered analysis with a full risk matrix — and not one named player. Not one event. Not one rule. Not one date.

A casual reader might take it for a broken report. I read it as evidence.

In sports analysis, people talk constantly about the error rate of conclusions. Very few talk about the error rate of the input. An empty input does not produce a wrong conclusion. It produces a temptation: to fill the gap with reasoning that sounds entirely plausible. In table tennis that temptation costs more than in many other sports, because table tennis is bound to the calendar — so tightly that a dateless file is unanalysable in principle, even if every other field were filled.

Tracing an extraction failure

At stage one, an empty object admits three explanations. The first: the upstream extraction step failed or returned an empty payload. The second: the source article was never an analytical item — perhaps a headline, a video caption, an image-only post — so there were no information points to extract. The third: a pipeline plumbing error, where stage one's output object was passed along without ever being populated.

All three point to the same action: don't analyse, fix the pipeline.

But one detail tilts me toward the first and third. The domain-label field was fully populated: table tennis. Which means the system successfully classified the article's domain. To classify a domain, a system must read the content. So that content existed somewhere in the chain, and vanished before it could be broken into information points. A genuinely empty source could not have been labelled with such precision.

There is a second detail. The time-sensitivity field was not overlooked by accident; it was explicitly marked as not assessed at stage one. A system that knows it has not assessed time sensitivity is a system aware of its own deficit. That awareness rarely appears in successful runs — it appears most often in runs where the input data never arrived.

This is the kind of reasoning I do in front of any data file: read the empty as well as the full. The grass surface is always beautiful. What has value lies beneath three layers of sediment and silence.

Applied here: the top layer of the file is the domain label — neat, tidy, credible. Beneath it sit three consecutive empty layers: no title, no source, no time. The final layer, the silent one, is the absence of any named person. That silent layer is what told me what had happened inside the pipeline.

Why table tennis will not forgive vagueness about time

There is a reason I place table tennis among the most date-sensitive sports, level with disciplines that run rolling points systems.

Professional table tennis rankings run on rolling points: points a player earns expire after a set period, usually calculated on a 52-week cycle. Today's ranking position is therefore the residue of points accumulated over the previous twelve months, and every week that passes strips part of that accumulation away. A player ranked very high may be standing on a block of points close to expiry; a player ranked lower may be at the very peak of an earning cycle. Look at a ranking table without looking at the clock and you see order, not pressure.

Then there is the position of a tournament within the year and within the larger cycle. The same event, played at the start of a season and at the end of it, is two different stories: different fitness, different accumulated match load, different hunger for points, and a different entry list. Seeding, draw, the chance of meeting a bad stylistic matchup in an early round — all are functions of the calendar.

In table tennis, time is also embedded in things that look unrelated. A change of ball, an adjustment to service rules, a new regulation on racket glue — with each of these, an entire generation of players must rebuild technique, and the timing of the announcement matters no less than its content. A change issued before an Olympic cycle and a change issued immediately after it produce two entirely different fates for the same group of athletes.

That is why a dateless file cannot be analysed. Not because it lacks data in some particular dimension, but because it lacks the time axis on which every other dimension hangs. I can read an article with no player names and still extract something about a trend. I cannot read an article with no date and say anything of weight about ranking, seeding or cycle.

This is a rule I set myself long ago: when the position of a piece of information on the time axis cannot be fixed, every conclusion drops to the level of hypothesis. There are no exceptions for cases that sound exciting.

Minimum sample: two mistakes and one rule

I have paid for ignoring that principle before.

In the summer of 2026 I spent weeks watching a youth-age tournament in Guangdong. I recorded fourteen matches of a fifteen-year-old left-back, rebuilt his passing map, and found an odd habit: he always cut into the half-space before passing, opening an angle the opposing defensive line never anticipated. I wrote a twelve-page report, complete with zone diagrams and coordinates, before I dared propose that a club sign him for a compensation fee of 150,000 yuan. He was signed. But the coach remained sceptical, because the system he preferred was zonal defending, in which that habit of cutting inside is a positional error.

The lesson I took away was not whether he was signed. It lay in the sample size: fourteen recorded matches, and a report many times longer than the sample.

A year later I made the opposite mistake.

In June 2026, in a group-stage match at a major tournament, a nineteen-year-old came on in the eighty-second minute and produced three progressive carries in nine minutes. I was swept up by the metric and wrote a piece declaring him the future of the wide-ball-carrying game. Veteran scouts laughed at it. He later barely developed, because of injury.

Arzani gave me a useful shock: the bigger the stage, the longer the shadow.

Since then I apply a hard rule: never assert anything about a player under twenty on fewer than five hundred minutes of competitive play. Five hundred minutes is the minimum threshold at which a sample begins to mean something — enough to see how that player performs when tired, when trailing, when the opponent has already read his game.

Then I realised that rule has a version for the writer. If a young player needs at least five hundred minutes to be asserted about, then an analysis needs at least one information point in order to exist. Not a good information point. Just a real one — verifiable, sourced, dated.

The file that morning contained zero information points.

Three layers of sediment and the cost of an empty cell

The method I use to assess a young athlete divides them into three layers. The first is learned technique: what a coach taught, observable, correctable. The second is habit formed by the training environment: how that player moves off the ball, how he reacts to instructions from the bench, how he positions himself in situations nobody drills. The third is instinct for reading the game — the thing that cannot be taught, visible only at moments when the player no longer has the strength to perform.

Reputation is noise. The signal lives at minute 70 — where people are too exhausted to pretend.

But all three layers need one thing to begin: observational data. Without a player's name, a match, a behaviour at a decisive moment, there is no layer to excavate. An archaeologist cannot dig sediment out of a plot of land that does not exist.

That was exactly the condition of the nine-dimension analysis that morning. Dimension two — player data and head-to-head records — needs at minimum a name and a ranking snapshot. Dimension three — event system and points rules — needs an event name and a date anchor. Dimension six — coaching staff and talent pipeline — needs a team and one staffing fact. Dimension nine — industry transmission — needs a commercial or policy trigger. Not one dimension met its minimum condition.

What is notable is that the report still carried a complete risk matrix. The six sporting risk groups — competitive, selection, generational gap, governance and public opinion, systemic, opponent — all returned blank cells. But a seventh risk, belonging to none of those groups, was flagged high: the risk to the integrity of the analytical process itself.

That risk states itself neatly: if a decision is taken on an empty object, any conclusion behind it can only be fabrication or hallucination. The report said so outright, and recommended halting the analysis chain, quarantining the result, and keeping it out of any downstream aggregation.

To an outsider, this must sound strange: how can an empty analysis be the correct result? The answer is that the value of an analysis lies neither in its length nor in the polish of its form. It lies in whether each assertion can be traced back to evidence. A three-thousand-word report with no evidence at all is not analysis; it is a formal structure stuffed with air. A report that states, accurately, that there is insufficient data and specifies what must be added is itself evidence — evidence about the state of the pipeline.

In thirty years of tracking academies and young talent, I have learned that what determines the quality of an analysis desk is not the number of pieces it publishes, but the number it refuses to publish.

The counterintuitive point: the real danger is the nearly-empty file

Stop there and the story sounds like a lesson in caution. I think it has one more layer, and that layer is the one worth telling.

An empty file incriminates itself. Anyone who looks knows it is unusable. It is a trap with no bait.

A nearly-empty file is different. It has nine minutes of a nineteen-year-old and three progressive carries; it has a beautiful metric, a beautiful moment, a rising name. It has enough material for a writer to build a fluent story, and nothing to make that writer hesitate. It is a trap with bait, and the bait is the feeling of holding evidence.

Put differently, the danger is not a pipeline returning zero. The danger is a pipeline returning a very small quantity that looks sufficient.

Modern table tennis falls into this trap with unusual ease, because the tempo of its information outruns the tempo of its samples. A tournament lasting a few days can generate dozens of beautiful moments, each enough for a headline, and almost none enough for a conclusion. Between those two things sits a gap, and that gap is always filled with language.

I have watched this repeat across sports. A young athlete beats a strong opponent in a big match, and within hours social media has a ready-made story about a new generation. Three months later that same athlete loses several matches in a row and the story disappears. Nobody is accountable for the story already written, because it was never recorded alongside its level of certainty.

That is why I believe in recording the empty. An empty result, written down with a date, an error code, and a link to the input object, becomes a marker. The next time a similar file passes through, the analyst will see the precedent and know the process permits them to stop. If the empty result lives only in someone's head and is then deleted, the next person will fill the gap with whatever sounds most plausible.

A fifteen-year-old does not need you to believe in him. He needs you to still be standing there when every camera has turned away.

The same goes for an empty data file. It does not need you to believe in it. It needs you to stand there and record that it is empty.

What has to happen next

Three things, in order of priority.

The first is a hard gate at the boundary between the two stages. If the information-points array returns empty, stage two must not run. That is far cheaper than discovering late that a long analysis was built on nothing.

The second is making publication date a mandatory stage-one field. No date, no analysis. For a sport where every metric hangs on the time axis, this is not a formality; it is the condition for analysis to mean anything.

The third is logging every null return, together with a description of the input object's state. Such records look useless until they become the only way to tell a random failure from a systemic one.

As for this article itself — it exists only to describe a file that contains nothing. That may sound wasteful. But in this trade I have seen many times that the thing most worth writing is not the moment someone shines, but the moment someone cannot, and whether we dare to say so.

An empty data file is not a failure of the analyst. It is a reminder that the analyst has kept his discipline.