Heatmaps and the Empty-Data Trap in Professional Table Tennis

HEATMAPS AND THE EMPTY-DATA TRAP IN PROFESSIONAL TABLE TENNIS

1. An Empty Data File in Shenzhen

"The first call came from a woman nobody names on the coaching bench."

I wrote that line into my notebook long ago, when I was a third-year sports science student sitting behind the bench of a women's team in Shenzhen. The person who called me that night was a head coach. She said I had looked at exactly the place the whole analysis room had overlooked. From then on I learned something that has stayed with me: the value of an analysis lies in what it reveals that others forgot, not in how many pages it fills.

Years later, I sat in a table tennis analysis room in Shenzhen, staring at a file with a full title, a table of contents, twelve statistical columns, and three colour charts — and not a single data point inside.

The assistant who sent it had been on the job seven months. He followed the process correctly. Pull the data, paste it into the sheet, export, send. The gap sat between two steps: between the system reporting "data returned" and the data actually having content. Nobody checked that gap, because the spreadsheet still looked tidy, the headings were still spelled correctly, and the formatting was still standard.

What chilled me was not the technical fault. It was the room's reaction. One person suggested "just estimate from the previous match." Another suggested "fill it in with the tournament average." Nobody suggested re-running the process.

I stayed quiet for about two minutes. Then I asked a question I have since asked many times, in many different analysis rooms: if we fill the gaps with invented numbers, is this still an analysis at all?

In modern professional table tennis, the answer is usually: yes, and nobody checks.

2. A Century, and a Digital Revolution

World table tennis is entering its hundredth-year anniversary. The first World Championships were held in London in 2026, when the sport was still played with celluloid balls, pimpled rubber on wooden blades, and all the data from a match lived inside the head of the person keeping score beside the table. In 2026, the World Team Championships return to London, exactly a century after that first edition. A circle closes in the very city where it began.

Between those two points, however, something changed far more than ball size, service rules, or the eleven-point system. That something is the volume of data generated around every ball.

When WTT was created and began running its event system in the early 2020s, table tennis shifted from a sport recorded by hand to a sport measured by machines. Ball-tracking systems appeared on major competition tables. Every serve was tagged. Every rally was broken into a sequence of events. The world ranking moved to a rolling 52-week mechanism, in which a result's points automatically expire after exactly one year — creating a pressure coaches describe with a very visual phrase: points-defence pressure.

Alongside that came a less discussed consequence. Once everything is recorded, people begin to believe that everything has been recorded.

In December 2026, Fan Zhendong, Ma Long and Chen Meng all withdrew from the world ranking. These were three Olympic champions, three names that had defined an entire generation of table tennis. The stated reasons revolved around mandatory-participation rules and withdrawal penalties, together with the physical and mental strain created by a packed calendar. The episode dominated international coverage for weeks.

To me, that was a data event before it was a sporting one. A measurement system can run smoothly, produce figures that are entirely valid in technical terms, and still lead to an outcome nobody inside that system wanted. The ranking was not wrong. The rules were not wrong. And yet the result satisfied no one.

That is the most precise definition of a distorted data system.

3. Three Layers of Reading a Table Tennis Match

Based on my experience watching matches, both live in the arena and on video in the analysis room, I have come to see that most arguments about data in table tennis stem from confusing three different layers.

The first layer is raw data. It is what the machine captures: the ball travels from point A to point B, the stroke takes 0.42 seconds, estimated speed 78 km/h, landing point 31 cm from the edge. This layer is objective, precise, and meaningless on its own.

The second layer is information. It appears when raw data is placed in a comparative context: this stroke happened on the ninth point of the fifth game, at 8-8, after a short serve. This layer begins to carry meaning, but it is still not enough to make a decision.

The third layer is tactical knowledge. It appears when information is placed inside a model of how the opponent makes decisions: this player, when trailing 8-9, tends to serve backspin toward the forehand side, then steps half a pace back to wait for the cross-court loop. This is the layer a coach needs before walking into the technical meeting.

The problem with modern table tennis is that the system produces a great deal of layer one, a fair amount of layer two, and very little layer three. Data providers have an economic incentive to thicken layer one: more metrics, more charts, more comparison tables. Layer three cannot be replicated that way, because it requires a specific human being to sit down, rewatch the footage, and offer a judgement that might be wrong.

4. Space First, Stroke Second

"In Croatia, I learned that a midfield does not chase the ball; it chases space."

I quote that line exactly once, because it is the only lesson I carried from football into table tennis. Everything else in table tennis has to be read in its own language.

In table tennis, a stroke is the final outcome of a decision about position. The player who wins the point is usually not the one who hits hardest, but the one who occupied the angle of the table before the ball arrived.

Take a very common situation in elite table tennis. A right-handed player stands shifted to the left, preparing for a forehand loop. The opponent serves short to the middle, slightly toward the left side. In raw data terms, this is an ordinary short serve. In tactical terms, it is a decision aimed at the blind spot of the movement system.

The receiver now has three options. Push short again, and surrender the initiative. Flick with the backhand, and open a rally in which he holds no angular advantage. Or pivot to the left for a forehand loop — a movement costing about 0.3 seconds and opening the entire half of the table behind him.

An analysis that merely counts how often the player chooses the third option will report that he pivoted 47 times in the match. An analysis that reads space will report that he pivoted 47 times, won only 19 of them, and that 14 of the losses came from exactly one serve location by the opponent.

The distance between those two analyses is the distance between data and knowledge.

Across many matches I have followed in the WTT system, the difference is never made by the speed of the loop. It is made by the first footwork step after the opponent contacts the ball. Today's leading players — Fan Zhendong and Wang Chuqin on the men's side, Sun Yingsha on the women's side — share one trait: their feet begin moving before the ball leaves the opponent's racket. That is a skill of reading intention, not of reading the ball.

No statistical table measures that interval. And that is precisely where modern table tennis analysis leaves a void.

5. The Heatmap and the Miscalibrated Compass

The heatmap has become the most widely displayed tool in table tennis analysis, and the most misleading one.

Its logic is simple. Every ball landing point is recorded, then coloured by density. Where the ball lands often is red; where it lands rarely is blue. Looking at the map, the viewer feels they are seeing tactics.

That feeling is wrong.

A heatmap tells you where the ball landed. It does not tell you who sent it there, under what circumstances, or why. A red zone in the left corner could signal a player who specialises in attacking the backhand side. It could equally signal a player who was pinned there all match and forced to block back into it. Two entirely opposite stories, one identical shade of red.

Worse, heatmaps are usually aggregated across a whole match or a whole tournament. That erases the single most important thing in table tennis: how behaviour changes with the score. A player can play completely differently at 5-5 than at 9-9. The heatmap merges all of it into one image, and that image describes a match that never took place.

In an internal analysis project I once worked on, we split a match into three score bands: below 7, from 7 to 9, and from 10 upward. The landing-point patterns in the high-pressure band differed markedly from the other two. Viewed only through the aggregate heatmap, that difference vanishes entirely.

A heatmap is not an analytical tool. It is a presentation tool. And a presentation tool used in place of an analytical tool produces a generation of decisions based on imagery rather than reasoning.

6. When the System Returns Zero

Back to the empty data file in Shenzhen.

What matters is that such failures are not rare. In any sports data system, a proportion of records will always be missing, misformatted, or mislabelled. The question is not how to drive that proportion to zero, but how to make the system recognise that it is holding a gap.

There are three ways to respond to a data gap, and they lead to three very different outcomes.

The first is to stop. Re-run the process, check the source, verify each record. This costs time and is often treated as sluggishness in an environment where a coach needs results before the next morning's technical meeting.

The second is to mark it. Leave the gap intact, label it clearly as "no data," and present the analysis with its holes explicitly mapped. This is honest, but it demands that the presenter be confident enough to say they do not know.

The third is to fill it. Use averages, use the previous match, use personal judgement, and present the result as though it were derived from data. This is the fastest, the smoothest, and the most dangerous.

In table tennis, the third approach has a specific name: faith in the model. Once a model has produced the right answer a few times, people begin to trust it as though it were a law. And when that model lacks data, they keep trusting it anyway.

A coach can look at a filled-in analysis sheet and make a substitution decision mid-game. That decision may be correct. But it is correct for a different reason than the one written on the sheet. And next time, when the sheet is empty again, people will fill it in again, decide again, and perhaps be right again. Until one time they are not.

7. The Execution Blind Spot

"The stands were empty, but I could still hear the coach shouting instructions at every metre."

I wrote that during a season without spectators, and it applies to table tennis in a different way. Inside a table tennis arena, the sound of racket on ball rings out clearly. But the thing that decides the outcome usually makes no sound at all.

It is the sound of the person on the coaching bench, signalling with a finger, a nod, or simply placing a water bottle on the table differently from last time. Those signals are recorded in no data system. And they change matches.

"A tactical wizard is not someone who sees more, but someone who looks where others forgot to."

The largest forgotten place in modern table tennis analysis is the interval between points. Those thirty-odd seconds contain almost the entire decision-making process of a match: the player wipes sweat, glances at the bench, receives a signal, recalculates, and steps into position to serve. No metric measures those thirty seconds, and because it cannot be measured, it is treated as though it does not exist.

"ENFP in the analysis room: finding inspiration in the driest numbers."

I still believe in data. I simply do not believe data knows what it is talking about. A complete statistical table from a table tennis match, read by someone with no understanding of space, pressure, or the interval between points, is nothing more than a nicely formatted document.

8. What Must Be Tested in London

When the World Team Championships return to London, a century after the first edition, thousands of metrics will be generated every day. There will be beautiful heatmaps. There will be detailed comparison tables. There will be predictive models built before the tournament begins.

The question I want to bring to London is not who will win. It is how many of those thousands of metrics will actually help a coach make a different decision than they would by simply watching the video?

And if the answer is very few, are we analysing table tennis, or decorating it?

A hundred years ago in London, points were recorded in pencil on paper. Today we record them with algorithms. The one thing unchanged is the person behind the coaching bench, who must still decide within thirty seconds, on incomplete information, and live with the outcome.

Perhaps that is the only data set that is never empty.