The Empty Table Tennis Dataset and the Cost of Filling Gaps with Guesswork

On the night of March 12, in my apartment in Shanghai, I opened a table tennis data file on my screen. Nine columns. Article title: blank. Source: blank. Article type: unclassified. Information points: an empty array, not a single atomic fact inside it. Exactly one field was populated — the domain label, two words: table tennis.

I stared at it for fifteen minutes. Fourteen years of writing about football before moving to table tennis had taught me to work with thin data. But a file where every field is empty, including the fields a machine fills automatically, belongs to a different category altogether. It is like walking into a press room and finding forty-seven seats with name tags on them and nobody sitting down. The name tags are still there. The people are not.

The gap is not on the diagram. It sits between people.

A genuine article, however thin, still leaves behind at least one name, one source, one timestamp, or a classified article-type field. Here, even the auto-populated field was empty. That means the text-reading step never ran — the failure sits at document fetch or format detection, upstream of analysis. Four causes, ranked by probability: a paywalled page, an anti-scraping block, a page rendered entirely in JavaScript, or a PDF or image that the parser could not read.

In table tennis, this kind of failure happens more often than people assume. The data for this sport lives scattered and fragmented. The ITTF ranking list updates on a fixed cycle, WTT results sit inside JavaScript-rendered pages, and the records for some continental events exist only as PDFs or photographs of paper score sheets. A text parser hits a scanned PDF and stops. It hits a paywall and stops. It hits an anti-scraping block and stops. What comes out is a nine-column file with eight empty columns.

In Vietnam the problem has an extra layer. The number of Vietnamese players reaching the main draw at WTT Grand Smash or WTT Champions events is very thin. Names such as Nguyen Khoa Dieu Khanh or Dinh Quang Linh appear in international result sheets often enough to be recognised, but not often enough to build a sample with statistical weight. A small sample means every conclusion is fragile. And that fragility is usually covered over with confident language.

There is a distinction most table tennis commentary skips: the difference between neutral and unknown. A dimension that has been examined and found clean is neutral. A dimension that has never been examined must be called unknown. The two get blended constantly, and the ITTF ranking list is where the blending is most visible.

The ITTF ranking system runs on a 52-week rolling window: the points from a result count for one year, then drop out of the total automatically. A player who does not compete for a cycle loses points, slides down the list, and gets described as being out of form. He is not out of form. He simply has not been measured. The 52-week system does not measure form; it measures presence — and it turns silence into a statement.

That is why I always split the question in two when reading a ranking list. First: how many matches has this player actually played inside the 52-week window? Second: of those, how many were against opponents from other associations? The second question is the one that answers the question about real level.

The win rate against players from outside the national association is the most important metric in table tennis analysis, and also the metric most dependent on having a specific named person. No name, no metric. A world ranking can be built out of domestic wins; but the moment a player steps into international competition, the gap between ranking and true win rate surfaces fast. The most important metric is always the first one to disappear when the data is empty.

In a normal analysis I would work through four layers: rally-length distribution, win rate on service points, footwork patterns of the pivot leg, and the way the coach calls instructions from the sideline. Each layer needs one minimum thing to exist — a name attached to a match with a date on it. Without the name, all four layers collapse together. An empty file cannot be rescued by analysing harder.

Meanwhile, the one populated field in the file is the least trustworthy one. The domain label says table tennis, but because every other field is empty, there is no way to verify whether that label reflects a real classification decision or a system default. At a draw ceremony, this is the equivalent of seeding a player because his name appeared on an entry form rather than because of his results.

The greatest danger is not the empty file. It is that a nine-dimension report with every cell reading insufficient information looks formally identical to a nine-dimension report with real content. Same number of sections, same table headers, same structure. The automated system downstream reads format, not meaning. A report with a complete skeleton can be read as a report with complete content — and that is the moment an empty signal becomes a wrong conclusion.

I have watched this mechanism operate elsewhere. At Paris 2026, Truls Moregardh's hexagonal blade generated thousands of articles within forty-eight hours. Most of them described the equipment; very few carried ball-trajectory data alongside it. The story about the blade moved faster than the data about the blade, and by the time the data arrived the story had already set.

The counterintuitive part: an empty file is more honest than a full file padded with memory. Table tennis analysis has a bad habit of converting memory into observation. This player turns slowly, she lacks nerve on deciding points — sentences like these sound like analysis but are really old impressions recycled for a new match. They are not obviously wrong, which is why they survive so long. And because they survive, they become the foundation for the next round of conclusions.

Writing insufficient information is harder than writing a conclusion. A conclusion gives a feeling of completion. A gap does not. But the gap was the only thing in that file that night which reflected reality accurately.

There is a boundary worth holding. When the input is zero, the risk of over-reading is greater than the risk of under-reading. I have seen people receive an empty document and immediately attach old stories to it — a selection controversy, a coaching change, a past scandal. The document never mentioned any of it. Memory did. That distinction matters, because it is the line between analysis and speculation.

The gap is not on the diagram. It sits between people.

The noise mechanism works the same way as in a transfer window. The volume of rumour rises; the volume of verifiable information does not. The ratio between those two quantities is the only useful filter. For table tennis, that filter has three questions: which outlet published it, on what date, and does the claim come with a fact attached. A piece of information that cannot answer all three is not information yet — it is raw material.

Table tennis readers today are submerged in a stream with high density and low resolution. They need a filter more than they need another opinion. That filter starts with separating three states: examined and problematic, examined and clean, never examined. Only the third state requires silence.

Next time you open a table tennis statistics sheet and see a blank cell, the task is to record that it is blank, record why it is blank, and then go find the original. A player who has not competed for a cycle is not a player in decline; he is a player who has not been measured. That distinction sounds small, but it determines everything that follows in the analysis — and it determines whether the analysis should exist at all.

The gap is not on the diagram. It sits between people. That holds even when nobody shows up in the press room. It holds even when the data file is empty. And it holds most of all at the moment someone is about to fill the gap with a story that sounds plausible.