A Table Tennis Analysis With Zero Data: Why I Wrote “Insufficient Information” Instead of Guessing

At 3:12 a.m. in Shenzhen, the nine-dimension table tennis analysis file came back from the data extraction desk: more than two thousand words, nine tables, forty-three content cells, and not one real data point inside. Every cell carried the same line: “N/A — insufficient information.” The technical table empty. The head-to-head table empty. The ranking table empty. The risk table empty, except for one bolded row: analytical-integrity risk.

I sat with it for fifteen minutes. Fifteen years in this trade and I had grown used to nights spent waiting for numbers, but never to a document that admitted it had nothing to say — and then said all nine parts anyway.

The youngest writer on the team would have deleted it and rebuilt it from memory. I filed it in a folder called “this month’s lesson.” The frightening thing in sports analysis is not missing data. It is a report that is complete, fluent, well-tabulated, and wrong.

To understand how a file like that reaches my desk, you need the workflow. At our newsroom, every deep table tennis analysis passes through two stages. Stage one extracts: read the source article, pull out discrete information points — which player, which event, which result, which source, which date. Stage two is the analysis itself: build models, compare metrics, form hypotheses. Stage two is only worth anything when stage one has returned at least one named entity and three verifiable information points.

That night, stage one returned an empty list. No title. No source. No article type. No players. In the notes field, instead of findings, it left an instruction: “derive the entities from the information points above.” A machine reminding itself what to do, then forgetting to do it.

In 2026, when the global calendar froze, I and a team of six built a database of 48,000 players across 32 leagues. We systematised every metric: distance covered per match, pressing intensity, point-win rate on serve, efficiency in the third-shot phase. Eight months with not a single match to write about. We chose to build foundations instead of headlines. That database has fed all of our transfer analysis ever since. The 48,000-player fortress: I did not save the world — I built a place where data could be safe. And it taught me that when the foundation is hollow, no beautiful house is anything more than a façade.

Reading that empty file closely, I found three signatures of a counterfeit analysis, and all three appeared at once.

First, instructions dressed as data. When the source field reads “judge from the source fields above,” and there are no source fields above, that is the fingerprint of a template emitted before the source article was ever parsed. In table tennis I use one test: if a metrics table names neither its unit nor its time window, it is not a metric, it is decoration. Based on my experience tracking matches, a 68% serve point-win rate only means something once you know how many points it was measured over, at which event, across which period. The same metric over three matches is an anecdote; over thirty matches it is a trend.

Second, an empty classification. The source came back as “unclassified.” For a system that has read thousands of articles, that label almost always means no text was read at all. Table tennis is a sport where the semantics of the article dictate how the numbers should be read: a piece about junior qualifying and a piece about a major final share a vocabulary but carry entirely different weights.

Third, and most serious: all nine analytical dimensions collapsed at once. Technique — nobody to dissect the serve, nobody to assess the third-shot phase. Ranking — the points-defence pressure under the WTT rolling 52-week mechanism cannot be computed without knowing how many points a player holds and when they expire. Events — the Olympic Games, the World Championships, the World Cup and the WTT tiers cannot be ranked without an event name. Competitive landscape — no tier map can be drawn when no association is named. Rules and governance — service-fault enforcement and rubber-thickness limits cannot be discussed without a related incident. Coaching and pipeline — generational transition cannot be assessed without ages. Risk — nothing to rank without a subject. Narrative — the gap between market expectation and reality cannot be measured when no expectation is stated. And the industry transmission chain — equipment, youth development, events, player commercial value — entirely blank.

One more professional detail: even the concept I use every day — international win rate — is meaningless here. Calculating it requires a named player, a list of opponents from other associations, and a time window. Assessing major-event consistency requires at least three editions of results. Assessing clutch performance requires point-win rates in deciding games. Not one of those three boxes was filled.

One detail stands out: all nine dimensions returned the same sentence, yet only one row was flagged high risk. That row is not about table tennis. It is about the writer. The greatest risk in sports analysis is not missing data; it is filling the gap with model memory and calling it analysis.

Here I have to state my professional position plainly. In 2026 I published a prediction the industry laughed at: I analysed a full national championship dataset and showed that a striker had an expected-goals figure of 14.8 but had scored only 8 — the unluckiest forward in the league. Veteran writers called it a mathematical farce. The next season he scored 27 and won the Golden Boot. The piece drew 1.2 million reads. But what I remember most is not the 27. It is that I had real data with which to be wrong, and was wrong in a verifiable way. Had I guessed from feeling that year, I would have been right in a way I could never prove — a kind of right that is more dangerous than being wrong. Numbers are the match’s love letter — learn to listen and you will see everything. Tonight the match said nothing at all, because it was never brought into the room.

Sports analytics is suffering from an illness few people name: surplus confidence, deficit evidence. We have more models than ever, more charts than ever, and a lower share of reports traceable to their origin than ever. A nine-dimension file full of “insufficient information” looks pathetic. It is still more honest than an eight-dimension report with numbers and one dimension invented.

Correlation is not causation — everyone knows the line, and almost nobody applies it when a deadline is squeezing. A player winning four straight matches after a rubber change does not mean the new rubber produced those four wins. You need to know which rubber, changed when, that the adaptation window usually runs three to six months, and who the opponents were in those four matches. Without those four data points, every equipment conclusion is belief decorated with jargon.

The paradox I want to put on the table: in a major-event season, when the public is thirsty for stories, the most honest analysis is usually the shortest one. Readers do not need us to fill the gaps. They need to know where the gaps are. Data does not answer your question. It teaches you to ask the right one.

I also have to audit myself. Before publishing anything, I write down a single piece of counter-evidence — one fact that, if true, would overturn my main argument. That night, my counter-evidence was the empty file itself. It was right. The claim “this article can be analysed” died on the first line. Trust is the only commodity this market misprices — until data corrects the price.

The signal I will track in the next cycle is not any player’s ranking. It is the share of “insufficient information” lines across the analyses we publish each month. If that share is zero across the whole system, we do not have a perfect process. We have a process that is fabricating. And if it hovers around a small but real level, I know the fortress is still standing — and still guarding the gate.