The Gaps Between the Numbers: When a Table Tennis Analysis System Returns Zero

The Gaps Between the Numbers: When a Table Tennis Analysis System Returns Zero

On a Tuesday morning, in the analysis room of a table tennis training centre in southern Guangzhou, I opened my machine and waited for the report from the previous day's session. The screen returned a flat grey frame. No athlete names. No serve-win percentage. No average rally length. No head-to-head table. Ninety minutes of footage, twenty-three variables to measure, and the result was a file full of meaningless abbreviations repeating steadily, like a record with the needle stuck in one groove.

I sat still for about three minutes before doing anything. In this profession, an empty data table is usually treated as the analyst's failure. But eighteen years of following the industry have taught me the opposite: on some days, the empty report is the most honest report there is. The problem is that almost nobody wants to read it, and even fewer want to publish it.

The Anatomy of an Empty Result

Before drawing any conclusion, I need to separate two things that the world of sports analysis habitually merges into one: a team with nothing worth saying, and a data-collection system with nothing it can say. These two situations look identical on screen, but their natures are worlds apart. The first is good news: we measured, and the emptiness reflects reality on the table. The second is bad news: we thought we had measured, but in fact the data pipeline broke somewhere between the camera and the spreadsheet.

In professional table tennis analysis, an empty result rarely comes from a single cause. It is usually the product of a chain of silent breaks. Perhaps the source text was empty — a recording corrupted in encoding, or a file truncated in transit. Perhaps the automatic filter was too aggressive — a ball-trajectory recognition algorithm wrongly discarded legitimate rallies because the lighting in the arena lacked contrast. Perhaps the manual labelling stage was skipped because the person in charge assumed the machine had done it all. Every link in that chain can break without making a sound.

What troubles me is not the technical failure. Technical failures can be fixed. What troubles me is the default human reaction to an empty data frame: to fill it in. In most analysis rooms I have walked through, an empty result gets handled in one of three ways. The first is to rerun the whole pipeline to see where the fault lies — the right way, but time-consuming, and nobody wants to explain to their superiors that today there is nothing. The second is to lower the measurement standard so as to obtain at least a few numbers — dangerous, because it manufactures fake data that looks like real data. The third, the most common, is to tell a story in place of a number.

The third way is the most frightening, because it leaves no trace. A coach receives an empty table, stays silent for a few seconds, then says: "The team moved poorly today." Nobody objects, because everyone "saw" it. But where did that "seeing" come from, if the very measurement system had failed to register the movement? It came from emotional memory, from a few rallies that caught the eye, from pre-existing bias. A conclusion with no data behind it is not a conclusion; it is a belief dressed in the clothes of statistics.

The Three Layers of an Analysis

When I discuss table tennis, I always divide everything into three layers, moving from the foundation up to the tip. The lowest layer is core technique — the mechanics of the stroke: racket angle, contact point, spin, footwork rhythm. The middle layer is tactics — how strokes are assembled into patterns: where the serve goes, which zone the receive places the ball into, who covers which area. The highest layer is the system — the training cycle, how people are selected, how fitness is allocated across a season.

An empty result can appear at any layer, and determining which layer it belongs to is the first step of any serious analysis. If the technique layer is empty, perhaps the camera did not have enough frames to capture the contact point. If the tactics layer is empty, perhaps the zone-labelling model was skewed. If the system layer is empty, perhaps the cycle data was lost between weeks. Mistaking the layer means being wrong from the root, and every conclusion built on top of it will lean the same way.

This is why I never begin an analysis session with an aggregate number. I begin by checking whether the pipeline is open. My first question is always: can I trust this table? Only when the answer is yes do I allow myself to read its contents.

The Technique Layer: When a Stroke Leaves No Numerical Trace

Table tennis is the sport with the greatest measurement paradox among all combat sports. A point at the elite level often ends after only three to five ball contacts. That means each point carries very little information, and a match of eleven points per game, up to seven games, produces a sample small enough to be worrying. A serve that wins a point outright says little if it happens only twice. A failed receive says little if the opponent just changed their racket rubber.

Precisely because the sample is small, every number in table tennis needs accompanying context, and that context often lies beyond the camera's ability to record. I once saw a statistics table stating that a player's serve-win rate had surged over the last three matches. Reading the number alone, anyone would think he had just improved his serve. But when I rewatched the footage, the truth lay elsewhere: the opponents in those three matches were all far-from-the-table defenders, who by nature have little ability to attack the receive. His serve had not changed. Only the opponents had changed.

This is where I must repeat a principle I have used for years: I trust data, but I trust the people behind the data even more — because both need to be coached. Table tennis data, at the technique layer, almost always needs a person standing behind it to read it with a human eye. A machine can count how many times the ball crossed the net, but the machine does not know that on the third time in the second game, the athlete's centre of gravity shifted slightly because his left ankle was tired.

The Tactics Layer: Formation, Gaps and the Language of Geometry

At the tactics layer, table tennis shifts from the individual stroke to the geometry of space. In all my writing, I rarely say "this player attacks well". I speak of the opening angle of the racket face against topspin, of the ball's trajectory dipping near the table edge, of the gap a movement leaves at the left corner. That is the true language of this sport, and also the language that raw data most often omits.

Picture a player attacking from the forehand position, then being pushed to the left. A gap forms at the right corner — not large, about a hand's span, but enough for the opponent to place the ball for a winner. A statistics table counting points lost will record that point in the "lost to attack" column. But the real cause does not lie in the final attacking stroke. It lies in the second step, two seconds earlier, when the player turned his body slightly in the wrong direction. The number records only the consequence; it does not record the cause.

I call these cracks in movement. There are cracks that do not show up on the tactical diagram, yet they tear open an entire campaign. A good coach does not read the points-lost column. They read the position of the feet in the frame before the ball leaves the opponent's racket. They read the distance between the two feet as the player transitions from defence to attack. They read the shoulder angle as the loop begins. These things almost never appear in an automatic data table, and that is why an empty result at the tactics layer is often accompanied by a match in which the human eye can plainly see the problem.

The System Layer: Pipelines, Components and Cracks

Moving up to the highest layer, we leave the racket and step into the machine room. A modern table tennis analysis system, whether at a small club or a national team, runs like a chain of links: recording, time synchronisation, ball-trajectory recognition, event labelling, index computation, interpretation, communication to the coaching staff. A single cracked link renders the entire chain behind it meaningless.

A system runs well only when the components inside it are not cracked. But the trouble with a link is that it breaks silently. A camera off by a few degrees reports no error. An over-tight filter reports no error. A skipped labelling stage reports no error. All of them stay silent, and the final product — the report table — still looks beautiful, fully formatted, properly presented. That is the most dangerous kind of failure in analysis: a failure whose result looks perfect.

It took me many years to learn to distrust a report table that is too perfect. If every cell is filled, every index reasonable, every trend clear, I start checking the pipeline again instead of believing the content. My experience is this: a real data table from elite table tennis always has a few empty cells, a few anomalous values, a few data points that contradict each other. That small messiness is the mark of an honest measurement process. Absolute cleanliness is usually the mark of a process that has been flattened.

The Rule of Three Field Checks

I have a rule that I apply to every trend, every conclusion, every finding: the rule of three field checks. A trend is only considered real when it appears at least three times across three different contexts, and each time is confirmed by a human eye alongside the number. One time can be luck. Two times can be coincidence. Three times begins to be a pattern.

This rule was born from the times data deceived me. In 2026, when competitions returned in empty arenas, I noted a strange phenomenon: one team repeatedly dominated possession yet lost more than it won. At first I blamed a skewed expected-goals index. But when I studied the footage closely, I realised the cause lay elsewhere: with no crowd pressure, opponents dared to push higher, and the home team lost its attacking build-up options. The empty stadiums of 2026 taught me that context is the most expensive thing that cannot be stored in a spreadsheet. Since then, I always record the context alongside every number I publish.

In table tennis, the rule of three matters even more because of the small sample. A player winning three serve points in one game says nothing. But if the same type of serve — same placement, same spin, same rhythm — wins points across three consecutive matches against three different opponents, then it is a real weapon. The difference between these two situations is the difference between noise and signal, and no algorithm can replace the analyst having to count to three themselves.

The Language of Geometry When Numbers Fall Silent

When data is empty, I return to the thing I trust most: the geometry of the table. A table tennis table is two metres seventy long, one metre fifty-two wide, seventy-six centimetres high. On that plane, every movement can be reduced to angle, trajectory and gap. A short serve sits neatly within the first section from the net. A forehand loop dips in the final third of the opponent's half. A pivot to the left opens a gap at the right corner about forty centimetres wide.

When I have no data, I draw. I redraw the table diagram, mark every landing point, connect them into trajectories, and look for which zone the opponent is exploiting. This method is slow, but it gives me something an automatic data table cannot: an understanding of space. When data stands still, I begin to read the gaps between the numbers again — and usually that gap is precisely where the match was decided.

There is a memory I still keep as a lesson of the craft. Years ago, while doing video analysis for a youth team, I once spent a whole evening redrawing ninety minutes of a match on a tactics board, cross-checking it against positional data. What I found did not lie in the aggregate numbers but in a small detail: the attacking players were drifting a few metres off the central axis, and the distances between the components of the system had stretched too far. I proposed narrowing those distances. In the return leg, the team won. The lesson I drew was not which number was right, but this: it is distance and running angle that make the difference, not the final result.

The Counter-Intuitive Angle: The Pressure to Have a Story

This is the part I want to spend the most time on, because it is the least discussed. In the sports media environment, the greatest pressure is not the pressure to be right. The greatest pressure is the pressure to have a story. A press conference cannot end with the sentence "we do not yet have enough data to conclude". A news bulletin cannot go on air with the headline "nothing noteworthy". An analysis piece cannot be a single line. So people fill it in. And when people fill it in with sentiment, what is produced is not analysis but a story adorned with a few numbers pulled from somewhere.

I once witnessed this at scale. At a World Cup, the entire media pack converged on the shock elimination of a major team. Everyone had an explanation. But when I spent three days rewatching every passage of play from the losing team's midfield, I saw a structural problem: two central midfielders kept moving into the same zone of space, causing the midfield to lose its structure, while the full-backs pushed high with nobody covering the space behind. That was a system crack, not an individual mistake. My piece on it was not widely read. But a club's coaching staff printed it out and pinned it to the meeting-room wall. Sometimes the value of an analysis does not lie in the number of reads.

The truth is, the public does not like cautious conclusions. They like a single cause, a figure to praise or to blame. But modern table tennis, like modern football, is a system. Reducing victory or defeat to a single cause denies the systemic nature of the sport. And when an analyst starts doing that to please the crowd, he has already ceased to be an analyst without realising it.

What an Empty Result Really Teaches Us

Back to the grey frame on that Tuesday morning. It took me nearly two hours to trace the fault. It turned out the pipeline was not broken at all. The recording had been corrupted in encoding at the transmission stage, turning the entire downstream data frame empty. When I reran it from the original file, the data appeared in full. But during those two hours, I had nearly done something I always tell my students not to do: tell a story about a match I had in fact never seen.

What I want to leave behind is not a technical warning. Technical matters can always be fixed. What I want to leave behind is a habit of thought: when the system returns zero, the first move is not to find a way to fill it in, but to determine whether that zero means "there is nothing" or "it has not yet been measured". These two are entirely different, and confusing them is the source of most wrong conclusions in sports analysis.

Takeaway

I still keep the habit of opening every report with a single question: can I trust this pipeline? If the answer is no, I close the file and go looking for raw data. If the answer is yes, I begin reading the content. In the regular season, as each round passes and each league table shifts by a few points, what I track is not who is leading, but which data pipelines are quietly cracking. Because I believe that, in the coming weeks, some conclusion will be drawn from an empty data frame that nobody noticed. The question for you, the reader: the last time you heard a confident claim about a table tennis match, did you ask yourself whether the speaker had actually seen the number — or was only seeing the gap between the numbers?