The Empty Column: The Limits of Table Tennis Analysis

At three in the morning the screen showed three values: 318, 47 and 0.

318 was the number of youth matches I had collected over four months across Asia's junior circuit. 47 was the number of technical variables I had built for that dataset, from short-serve rate to rally-length distribution. And 0 was the number of values sitting in column 47 — the column that mattered most, the one measuring third-ball win rate for players under nineteen.

The column was empty because three of the four local data providers recorded only game scores, not point outcomes. A 3-1 win says nothing about how often a player wins on the third ball. I sat in front of a spreadsheet full of names, dates and scorelines, holding none of what I actually needed.

That night I wrote a four-page report. Its conclusion fit in one sentence: insufficient data to conclude.

Four pages for one sentence. It is the hardest report I have written in more than twenty years in this trade, and the one I have kept longest.

A season with no room for emptiness

Table tennis has no season in the football sense. The WTT system runs on a 52-week cycle: each event's points last exactly one year, the world ranking counts a player's best eight results, and a player can drop three places in a week without losing a match — simply because his best eight results aged out together. Points-defence pressure, therefore, does not come from the opponent across the table. It comes from the calendar.

For a writer, that structure imposes a particular rhythm. At least one WTT event closes every week; each event produces dozens of matches; each match produces hundreds of points. Readers follow every match and they want tactical signals before those signals become headlines: why this player lost the serve in game four, why her third-ball win rate fell twelve percentage points against the previous event, why a nineteen-year-old won four of five deciding games while his senior lost three.

That demand is reasonable. It is also where this trade slips most often.

The Chinese market, where I work, is where that demand bites hardest. One Grand Smash generates hundreds of stories a day, and speed is always placed ahead of accuracy.

At Paris 2026, table tennis awarded five gold medals. Fan Zhendong completed the career Grand Slam in men's singles. Chen Meng defended her women's singles title after Tokyo. Wang Chuqin and Sun Yingsha won the mixed doubles. Those three facts can be verified with a single search. But around them, in that same week, thousands of articles asserted things nobody can verify: that player A is in decline, that player B has been decoded, that style C is obsolete. Most of those assertions had not one line of data behind them.

A ranking is a fact. A form forecast is an inference. The two get written into each other, and when they do, the prose reads smoothly — and the error becomes very hard to spot.

Three tiers and four gates

I sort everything I read into three tiers, and I never let the tiers mix inside a single paragraph.

Tier one is verified fact: scorelines, match dates, head-to-head results, ranking points, entry lists. This tier can still be wrong, but it is wrong in a fixable way — look it up and it resolves.

Tier two is grounded inference: a conclusion drawn from tier one through a logical step that can be written out in words. A player wins 68% of her deciding games over twelve months, the leading group averages 55%, so there is a basis for saying her late-match pressure tolerance sits in the upper band. The logical step is the assumption that a deciding game is the highest-pressure situation available — an assumption stated out loud, and one the reader is entitled to reject.

Tier three is hypothesis awaiting follow-up: things I believe but cannot yet assert. Tier three has to be labelled clearly and must carry a date — which event do we check this against?

These three tiers are the only barrier between an analysis and an advertisement for the writer's imagination.

Four gates apply to every dataset before I allow myself to write. Gate one: at least one named entity — a player, an event, an association. Without a name there is nothing to analyse. Gate two: at least three independent information points. One point is a brief. Two points are a suspicion. Three points begin to form a hypothesis. Gate three: a non-empty source field — I have to know whether the information came from an official release, a press conference, event data, or an unnamed social account. Gate four: an absolute date. "Last week" is not a date; "13 August" is. In a 52-week rolling ranking, being one week off means being wrong about the entire points-defence picture.

On the night column 47 was empty, my dataset failed gate two. I had 318 matches but only one information point that actually answered the question I was asking. The other three hundred and seventeen rows were correct and irrelevant. Volume does not substitute for relevance.

Two traps

The first trap is the fame filter. When data is missing, writers substitute reputation. A famous player performing badly is explained as "a psychological crisis"; an unknown player performing well is explained as "finding form". Both are answers that require no data, and are therefore answers that cannot be tested. The fame filter runs so smoothly that readers do not notice they have been handed a conclusion instead of a measurement.

The second trap is old memory. When a data field is blank, the brain fills it with the most recent value it saw. In table tennis this is more dangerous than in most sports, because WTT rankings expire week by week. A seventh place I remember from last season may be fourteenth today, and an entire draw analysis collapses with it. I have learned to write "pending verification" into any cell whose provenance I cannot trace, even when it makes the sentence uglier.

Source tiering is the least discussed step and the most decisive. I rank sources in four tiers. Tier one is a direct statement from someone involved, or an official release from the organiser. Tier two is journalism with an editorial check and a reporter on site. Tier three is second-hand reporting. Tier four is anonymous social content. Information about an injury is allowed into my copy only at tier one or tier two. Information about internal team tension almost never reaches tier one — which is why most of it should not be written at all.

Four metric families and one gap

Data is only bone; the story of a match is flesh. I hold the scalpel slowly.

In table tennis, four metric families separate one player from another. The first is the third ball — win rate when the server takes the initiative inside the first three contacts. The second is receive error rate, split by spin type. The third is rally-length distribution: a player who wins 70% of points ending inside four contacts but only 40% of points beyond seven is two different people, and any opponent who reads that will stretch the match. The fourth is deciding-game win rate — the only direct measure of what people loosely call nerve.

One metric sits outside those four and governs long-term value: win rate against players from other associations. At national-team level it separates players who are good domestically from players who are good internationally. Plenty of players have glittering domestic records and average international win rates, and the gap between those two values is the entire content of a serious piece of analysis.

Across more than a decade of watching junior matches in Shanghai, Guangzhou and Taipei, one observation repeats: young players win with the thing they are best at, and lose with the thing nobody has ever measured in them. A seventeen-year-old with a heavy forehand loop usually wins the first three games of everything, then loses the last two when the opponent changes the serve pattern. The scoreline does not record that moment. Only point-level data does.

The rough gem shows itself on the third ball under pressure, not on the beautiful loop in front of a full stand.

The other side of certainty

Here I have to argue against myself.

The whole system above works well until it meets a situation where the data does not exist — and in table tennis that happens more often than outsiders assume. Junior events have no point-level statistics. Academies do not publish physical data. Injuries are concealed until they cannot be. At that point the natural reflex of a professional is to fill the gap with experience, with intuition, with "a feeling built over twenty years".

Intuition is not wrong. Intuition presented as data is wrong.

The most dangerous analyst does not sit in the group that gets things wrong. He sits in the group that always has an answer — because for him an empty dataset is a stage, not a stop sign. He will talk about a new geological layer in the sport, about a generational shift in a national team, about a crisis of identity in a playing style — all of which may be true, and none of which can be tested. The fluency of the prose conceals the emptiness of the evidence.

I have been in that seat. In my third year on the job I wrote a piece about a young player based on two matches watched in person and one video clip. The piece was good. The piece was wrong. The error was not in the conclusion but in my failure to state that I had only two matches.

Losing a season is not losing a site; put the map down and think again.

What I leave behind

My dataset now has column 47 — incomplete, but present. Four months of collection, three providers changing formats, one new scoring system under trial at two junior events. I still will not write a conclusion about the under-19 group, and I probably should not until the next 52-week cycle closes.

What I want to leave behind is not a method. It is a question for the people writing about table tennis: when a piece of analysis cannot muster three verifiable data points, is the writer serving the reader or serving his own sensation?

Do not ask what that player does with the ball. Ask what he does on the second ball — and ask the writer what he used to verify the answer.