One Empty Data File and the Discipline of Numbers in Modern Table Tennis

At 2:07 in the morning of 12 March, the wall clock in my Shanghai office ticked past the hour. A text file had just landed in the analysis system: a first-stage deconstruction of a table tennis article. I opened it. Nine fields. Eight said N/A or were blank. The ninth said exactly one thing: table_tennis.

No title. No source. No article type. No information points. No named entity. No timestamp. No source-quality assessment.

Every revolution begins with a number someone forgot on a desk. That night, the forgotten number was zero.

My desk has three monitors. The left one runs a WTT Star Contender qualifier, cropped tight to the table edge so I can see the receiver's feet. The middle one holds the ITTF equipment catalogue, where every legal rubber carries its own code. The right one is the ranking ledger — the page I open more often than any news site.

My workflow starts with a deconstruction file. It must return nine fields: title, source, type, information points, the author's core viewpoint, the entities involved, time sensitivity, source quality, and domain label. Those nine fields are the skeleton. Without them, everything I write afterwards is literature, not analysis.

That night the system returned one field.

First I checked the connection. Then the system log. Then I sat still for about four minutes, because I knew exactly what temptation was waiting at the corner of the room: the temptation to write a beautiful table tennis piece out of what I already knew, and staple it to an empty file.

I have been in this trade long enough to know how dangerous that temptation is. Table tennis is a sport where a writer can invent a very specific, very technical detail and nobody can check it: sponge hardness, blade ply count, a closed-door "technical overhaul". Those details sound professional. They are the easiest lies to live with in sports media.

Some context on the ledger I open every day.

Professional table tennis runs on the WTT system, the events arm of the International Table Tennis Federation. Events are tiered: Grand Smash, WTT Finals, WTT Champions, Star Contender, Contender, plus the three biggest prizes of a career — the Olympic Games, the World Championships and the World Cup. The points gap between tiers is huge: a Grand Smash champion takes 2,000 points, a WTT Finals champion 1,500, a Champions champion 1,000, Star Contender 600, Contender 400. The three majors carry their own scales and their own weight in national selection calculations.

The decisive mechanism sits elsewhere: points expire after 52 weeks. Whatever you earned at an event is deducted at the corresponding date the following year. The ranking is therefore not a photograph; it is a debt book. Every player at every position carries a repayment due within weeks or months.

That is why a table tennis article without a date is an article that cannot be quantified. The same 6,000 points, at two different moments, describe two completely different states: one player defending 3,000 points in six weeks, another defending only 400 in the same window. Same total, similar rank, incomparable pressure.

Back to the empty file.

The only surviving field was the domain label. In any data pipeline, the last fields to survive are those the system generated, not those the text generated. A table_tennis label is usually assigned from the URL, the source tag, or the metadata. It needs no sentence from the article. It survives even when the article does not.

In table tennis there is a very concrete equivalent. When a match video file corrupts, what remains in the tournament software is the scoreline. What is lost is the first-bounce placement of the serve, the timing of the receiver's split step, the depth of the third-ball loop. The scoreline is what the system records. Structure is what only a human watching can record.

Scorelines are cheap. Structure is expensive. And analysis only earns its keep when it pays for the expensive part.

That is the first lesson an empty file taught me.

The second lies in a technical defect I consider more serious: the "entities to identify" field was instructed to derive from "the information points above", while the information point list was empty. A closed loop. Field A depended on field B, and field B did not exist.

In table tennis, the closed loop is the mistake I meet most often in tactical reviews. Someone builds a receive plan on the assumption the opponent serves sidespin, while the only data supporting "the opponent serves sidespin" is the very matches the plan was built on. Circular reasoning produces an intense feeling of certainty, because every step has data, and all the data comes from one source.

I have made exactly this mistake. I once wrote that a player's backhand was declining with age, based on his last four matches. All four were against left-handers, and all four featured short serves to the middle of the table that forced him to pivot. I measured age while I was actually measuring serve placement.

The third lesson is the timestamp. The deconstruction file recorded time sensitivity as unassessed. In many sports that is a minor omission. In table tennis it invalidates everything downstream.

Because of the 52-week roll-off, the same fact is worth different things depending on the month. A player who won a Star Contender in November enters March with 600 points about to expire. Another who reached the same final in May carries a debt that is still distant. Write about both in March without a date and you have manufactured a meaningless comparison and presented it as a meaningful one.

There is another layer: selection windows. Every national association sets cut-off dates for Olympic, World Championship and World Cup rosters. A result three weeks before the cut-off carries different weight from the same result three weeks after. Remove the date and this entire layer evaporates, leaving a drifting technical description.

The fourth lesson — and this is where I want to spend the most words — is equipment.

In table tennis, equipment is the most checkable thing in the sport and the most frequently fabricated. ITTF rules cap total racket covering thickness at 4mm, with the topsheet no more than 2mm. From 1 September 2026 the volatile-solvent limit was driven to zero, ending the speed-gluing era. The ball moved from 38mm to 40mm in 2026 and to the 40+ plastic ball in 2026. From October 2026 the two sides of a racket must be clearly different colours, one black and one bright, ending the two-red era.

Every legal rubber has a code in the ITTF equipment list. Which means every claim about rubber, blade or hardness can be checked in thirty seconds.

And yet most of what readers encounter about table tennis equipment carries no such code.

My self-imposed rule since 2026 is simple: I do not write a single sentence about a rubber unless I have either a photograph of the code on the packaging or the corresponding line in the ITTF equipment list. Without one of those two, I do not write it. Even when the source sounds certain.

The reason is not pedantry. The reason is that equipment is the cheapest place to lie, precisely because it is easy to verify. Writers know very few readers will spend thirty seconds looking up a code. The hook is not the truth; the hook is the feeling that the writer was inside the training hall.

The fifth lesson takes me back to what I actually believe: the micro level.

The first three seconds of a match do not lie. Everything after that is how we talk ourselves into believing.

In table tennis those three seconds live inside a single serve. The toss rises at least 16cm, above the table surface and behind the end line. Contact. The ball bounces first on the server's own half. Then it crosses. In that window I can count four things and write them down, and none of the four appears on a broadcast graphic.

First, toss height against the 16cm threshold. A server who deliberately tosses lower than the rule allows usually hides more spin, and umpires do not always catch it from a side angle.

Second, contact point relative to the top of the net. High contact allows deep backspin; low contact forces a fast, flat serve.

Third, the first bounce on the server's own half. I divide that half into six zones and count frequency across ten consecutive serves. Those six zones tell me what kind of ball the server is actually producing, not the kind he claims to be producing.

Fourth — and this is the one almost nobody counts — the receiver's foot position at the instant the ball leaves the racket. The split step lands a tenth of a second earlier or later than the ball's descent. Earlier, and the receiver has read the spin and will attack first. Later, and the receiver has already lost before contact.

What the audience sees on screen is not the real match. The real match happens in the gaps the camera leaves behind.

The broadcast camera follows the ball. When the ball crosses to the other half, the frame cuts to the receiver. In that cut, what is left at the edge of the frame is the server's recovery step — and that is the data that decides who wins the fourth ball.

The sixth lesson is one I write reluctantly: I was wrong.

Last year I published a conclusion that a player's backhand point-win rate had fallen after he changed blades. I had data. I had a count. I had a photograph of the new blade. The argument was airtight.

Three months later I re-coded three matches using the same method and found the error elsewhere. The backhand rate had not fallen because of the blade. It fell because all three opponents served short to the middle of the table roughly 40 percent more often than their own baseline, forcing him to use footwork instead of pivoting, so he ended up playing backhands from a passive position.

I was wrong. I wrote those words in the correction.

"I was wrong" is not a confession. It is a measuring instrument. An analyst without that instrument has no way to calibrate himself, and after a few years he will believe everything he has ever written, including the wrong parts.

Based on my experience tracking matches over many years, I have settled on one rule: each tactical conclusion should carry exactly one lever, and every secondary finding should be pushed into a footnote or dropped. An article stuffed with ten findings has none that can be checked. An article with one finding invites the reader to verify it — and if they find I was wrong, I can still correct it.

Now to what I think is the blind spot of the whole trade.

The popular belief in sports media is that more data means more certainty. In practice it runs the other way. Table tennis does not lack data; it lacks traceable data. Hundreds of articles a day carry micro-details no one can verify: sponge hardness, blade ply count, a closed training session, a conversation with a coach that only the writer attended.

Those details do not make a piece more credible. They make it harder to catch.

That is why I chose to publish an empty file. Not to boast that my system would rather leave a blank than invent. Rather to say that in an environment where anyone can produce a technical detail that sounds real, the highest-value act available to an analyst is to state clearly that he has nothing.

There is one filter I deliberately avoid: the fame filter. When data is missing, people gravitate to the most famous person in the room, because writing about the famous always feels correct. But the most famous person is not the one deciding the match. In table tennis the decider is usually the receiver in the second phase of the seventh game — no name on the scoreboard, half a second at the edge of the frame.

A table tennis article today with no date is a rumour wearing good clothes. One with a date but no source is a rumour with an expiry date. One with a date, a source, one finding, and one admission of ignorance is the only kind still usable six months later.

I am not arguing that emotional commentary should disappear. It has its place, and it needs no data to exist. The problem is that the two genres are being blended, and readers are not told which one they are reading.

What I want to leave for the next match is concrete and testable.

Next time you read a table tennis analysis, count two things. Count the numbers in it. Then count how many of those numbers are tied to a specific date and a specific source. If the second count is lower than the first, the piece is simulating analysis rather than performing it.

As for me, from the next match I will add a column to the notebook: the exact moment I measured the receiver's split step. A tenth of a second earlier than the ball's descent — and I will record the date I saw it. If I later conclude wrongly, I want to know precisely which day I was wrong.

Data does not replace an old coach's intuition. It only hands him a more accurate map. And an empty map, honestly labelled, remains more useful than one drawn from imagination.