When the Data Sheet Goes Blank: Reading a Table Tennis Match Through Angles, Trajectories and Gaps

My tablet lit up at 19:40, and the screen showed a blank sheet. Twenty-four cells, not a single number, and a familiar status line at the bottom: insufficient information. Four metres away, a round-of-16 match at a WTT Star Contender event was running as normal, the scoreboard ticking through 11-9, 9-11, 11-7. The stands were loud. The umpire still called service faults. The two players still wiped sweat at the change of ends. Only our analysis system was silent.

For three games in the middle of that match, I put the sheet away and did what I had done instinctively eighteen years earlier as a video analysis assistant for a youth team: I watched. I measured with my eyes the contact height of the red-shirted player's forehand loop and saw it drop roughly 8 cm from the first game, from around net height to just below the net tape. I counted the rhythm of the first footstep after short serves and found it slower by nearly a tenth of a second. I estimated the gap at the elbow zone, where the forehand and backhand wings meet, and saw it widen by about a hand's width whenever the opponent pushed long to the backhand. Not one of those three things had a cell on the sheet.

When the data stands still, I start reading the gaps between the numbers again. That night, the gap was as wide as a whole match.

A data layer thinner than it looks

Professional table tennis today runs on a data layer far thinner than the feeling it creates. World Table Tennis was founded in 2026 and launched its new event system in 2026, tiered from Grand Smash down through Star Contender, Contender and Feeder. At the top tier, according to World Table Tennis figures published for the 2026 season, the Singapore Smash prize pool sits at USD 1.5 million, bundled with a data package covering rally length, service placement, ball speed and spin classification. That is the visible part.

The invisible part is the production chain behind it: cameras, calibration of the coordinate frame, manual or semi-automated tagging, the data feed, and only then the dashboard that a coaching staff member like me actually sees. That chain breaks at any link. And when it breaks, the dashboard does not light up red. It shows empty cells that look very much like a clean report.

What cameras record only means something when someone reads it. In the Paris 2026 Olympic men's singles final, Fan Zhendong beat Truls Moregard 4-1; in the women's singles final that same cycle, Chen Meng beat Sun Yingsha 4-2, according to the official results published by the International Olympic Committee. Both matches were captured at the highest level of detail, and both showed the same thing: the distance between two players is not in the total score, it is in who controls the rhythm of the first three points after each change of ends.

I went through a lesson of the same kind once, in a different sport. In 2026, working as an analysis assistant for an U19 side, I counted only 2 successful presses out of 18 in the opponent's defensive third. I redrew the entire match on the board, cross-checked against positional data, and found the forwards were running 4 to 6 metres off the central axis. My only recommendation was to narrow the distance between the two wide midfielders from 28 metres to 22 metres. In the return leg, the team won 2-0, and nobody mentioned the number 22 in a single match report.

Three years later, when competitions returned in empty stadiums, I noticed something odd: six matches with no crowd, four with over 60 percent possession, but only one win. At first I blamed expected-goals distortion. Watching the footage back, I saw opponents pushing higher because they no longer felt the pressure of a crowd, and our build-up lost its shape. I built a 45-metric comparison between the crowd season and the pandemic season and concluded that the long-range pressing indicator was off by 32 percent. The empty stadiums of 2026 taught me that context is the most expensive thing a spreadsheet cannot store. Table tennis is the same, just on a different scale: a hall without applause changes whether a player dares to attack the third ball.

Three failure layers behind a blank sheet

When the sheet goes blank, the fault is never in one place. It sits in three layers, and each layer can claim innocence.

At the capture layer, cameras can lose calibration when arena lighting shifts between games, or when the table is moved a few centimetres during set-up. In table tennis that margin is not small: a 40 mm ball leaving the racket at over 100 km/h travels more than two and a half metres, and a misaligned coordinate frame misclassifies every placement.

At the classification layer, the system labels a serve as short, long, topspin, backspin or no-spin. The labeller is human, or a model trained on humans. A sidespin serve with a touch of topspin slides easily into a near-enough category, and that error compounds across a match.

The interpretation layer is where the real danger lives. When a cell is empty, a young analyst tends to fill it with an assumption, because assumptions are always available. A system only works when the parts inside it are not cracked. A crack in the interpretation layer does not turn the sheet red; it simply makes the conclusion wrong, very politely.

The percentage, and the structure behind it

Last season I tracked a player whose third-ball attack win rate on his own serve was 62 percent. That number is handsome enough for a slide. Split by service type, the picture turns: on short serves to the backhand, the win rate is 71 percent; on long serves to the forehand, it falls to 38 percent. The average of those two zones is 62 percent, and nobody saw a crack.

I now ask the same question of every aggregate: which cells was it assembled from, and do those cells share one condition? In table tennis, that condition involves at least four variables: service type, the opponent's stance, the physical state at that moment, and the score situation. A 62 percent figure that merges all four is arithmetically correct and almost useless for coaching.

Numbers do not lie, but they know how to stay silent about the part that matters most.

Geometry when the numbers are gone

On that night without data, I went back to four measurements the human eye can still take, even with a larger error margin than a camera.

Contact height is the cheapest and most sensitive measurement. When a player tires or loses confidence, the contact point drifts down, usually below net level. That height determines how much safety margin the ball has over the net, and therefore whether the player dares to strike through the ball or has to brush up on it. An 8 cm drop sounds small, but it is enough to turn a powerful loop into a defensive push.

The elbow zone is the second measurement, and the most neglected. It is the area where a player must decide in roughly a third of a second whether to rotate into a forehand or pull back to a backhand. A wider elbow gap means the footwork is off-axis, and an opponent only has to push the ball into that zone to make the whole defensive structure collapse.

First-step rhythm, measured as the time between the opponent's contact and the player's foot leaving the floor, is an earlier signal than muscle fatigue. It does not appear in any physical metric I have ever read.

Shoulder rotation range, measured as the angle between the shoulder axis and the hip axis at impact, closes the set. That angle determines the spin rate of the ball, and it also indicates whether a player still dares to hit with the whole body or has switched to hitting with the arm alone.

These four measurements do not replace data. They keep me from standing around waiting for a spreadsheet. What matters is that all four are geometry, not probability. Elbow gap, contact height, shoulder-to-hip angle, footfall rhythm: this is the language any tracking system must ultimately reduce to, and it is also the language a player understands immediately when you say it out loud during a timeout. The first time I saw collective movement as a piece of music, with coaching as the tuning of each note, was in conversations built entirely out of that geometry.

The cost of tracking too much

The paradox in this profession is rarely stated out loud: the more detailed the tracking system, the greater the risk of optimising the wrong variable. Give coaches rally-length data and they start encouraging players to extend rallies. Give them ball-speed data and they start encouraging players to hit harder. Both can be right, and both can destroy the structure of a playing style.

In that 2026 U19 match, my recommendation was not to run more, but to run narrower: to narrow the wide midfielders from 28 metres to 22 metres. That was a geometric fix, not a data-volume fix. Had I owned three more running-distance metrics at the time, I might well have recommended something worse.

In table tennis, the equivalent variable is the distance between the forehand-corner starting position and the elbow intersection point. Narrowing that distance does not make a player run more. It makes the player decide less, and in a sport played at this speed, deciding less is a bigger advantage than running faster.

Three field verifications

I apply one simple rule to every new technical trend, and I believe it should be the minimum standard in this profession: something can only be called a foundation once it has passed through three different contexts.

The lightest context is training, where pressure is low and errors are forgiven. The middle context is an event with weaker opponents, where the technique must withstand real pace. The heaviest context is a match with scoreboard pressure, where decisions are broken by tension.

Many new techniques survive the light context, occasionally the middle one, and vanish under the heavy one. Three field verifications are how you separate a fashion from a foundation. Living in an environment where new techniques appear every day, I have to keep this rule, not because I am conservative, but because I have counted too many beautiful things that disappeared after a single season.

The blind spot: no warning flag read as no risk

This is the part I want to state most plainly.

In any analysis system, missing data has never been the most dangerous fault. The most dangerous fault is missing data that looks like clean data. A dashboard without a red flag gets read as a dashboard without a problem. No one in the decision chain is trained to tell those two apart, and so the whole chain drifts.

There are cracks that never appear on a tactical diagram, and they tear a whole campaign apart. Those cracks usually sit exactly where the report says "insufficient information", and get skimmed past as an administrative detail.

The reverse is also true, and it is said less often: sometimes the correct output of an analysis session is a single line reading "insufficient information". Not every match contains a readable signal. Some matches are so lopsided, or so cautious, that every pattern is just noise. Our profession has no language for that honesty, so instead of saying it, people invent. Inventing a plausible trend is always easier than admitting there was nothing to read last night.

The execution consequence is very concrete. A coach who receives a clean report becomes more confident during a timeout, calls a tactic based on an unverified pattern, and loses three straight points before realising the pattern never existed. I have watched this happen, and I have also been the person who wrote the report that caused it.

I trust data, but I trust the people behind the data more, because both need to be coached.

What to verify in the next match

For the next fixture in the regular-season calendar, I will watch one narrow window: the first three points after every timeout, and the first three points after every change of ends. Those are the moments when structure breaks and rebuilds, where footfall rhythm, contact height and elbow gap tell the truth better than any aggregate. Based on my experience following matches across the WTT system and domestic events, most matches are decided in exactly those moments, not across the whole match.

The question I leave for myself: if the data sheet goes blank again, will I dare to write one line in the report reading "insufficient information", instead of filling it with a story that sounds reasonable?