Video Review in Professional Table Tennis: The 'Clear and Obvious' Standard Has No Ruler

The score was 10-10 in the seventh game. The player on the left side of the table bent forward, rotated her shoulder, and tossed the ball. The umpire raised an arm: service fault, ball hidden. The arena roared once and then went quiet. The server placed a hand on the scoreboard and requested a video review.

Three minutes later, the review panel announced the original decision stood. Two more points, and the match was over. The winner bowed to the crowd. The loser walked straight out the side door and never looked up at the big screen.

I was sitting in the sixth row, behind the technical area, holding the notebook and tracking sheet I had used all season. What made me stop was not the final decision. It was the arena. When the big screen replayed the slow-motion angle, about four thousand people looked up, held their breath, and failed — together — to point to the exact moment the ball left her palm. The people sitting courtside, less than three metres from the service, failed too.

Then a decision was made, and it was written into the record like every other decision.

All season long I pursued a question that sounds simple: what happens to a video review system built on a "clear and obvious error" standard when the decisive moment is not clear to anyone in the hall?

Twenty-five years ago, table tennis began a sequence of rule changes, and each one redistributed advantage between groups of players. The ball grew from 38 mm to 40 mm, reducing speed and spin and lengthening rallies. The 11-point game replaced 21-point play, raising the value of each individual point and turning the opening points of a game into tactical terrain. The 2026 ban on hiding the ball during service stripped a whole generation of classic servers of their weapon. The speed-glue ban reshaped racket structures. The plastic ball replaced celluloid in 2026 and cut spin once more.

None of those changes was neutral. Neither is video review. It favours the player who reads situations faster, the coach who manages challenge budgets well, and anyone who can control the tempo of the three empty minutes a review creates.

At Paris 2026, Fan Zhendong won men's singles gold and Chen Meng won women's singles gold. Sun Yingsha has held the world No.1 ranking for most of the period since, alongside Wang Chuqin on the men's side. Ma Long, singles champion at Rio 2026 and Tokyo 2026, closed his Olympic career with a team gold. On the other side of the net, Tomokazu Harimoto, Hugo Calderano and Félix Lebrun have become regulars in the quarter-finals.

At that level, a single point can decide an entire sponsorship cycle. And a single umpiring decision can decide a point.

Context: a process that unfolds in public

Football's video assistant referee operates in a closed room hundreds of kilometres from the stadium, with dozens of synchronised camera angles and an algorithmically drawn line. Table tennis has none of that. In table tennis, the process happens at the table, in front of the crowd, in real time.

A player raises a hand. The umpire stops play. The umpire, the assistant umpire and the chief referee gather around a monitor at the edge of the technical area. The footage is shown on the arena's big screen so the crowd can watch. The panel can call up additional angles, slow motion, freeze frames. The chief referee then announces the outcome over a microphone.

The list of reviewable situations is narrow. The ball touching the edge of the table or the side edge. A service that is hidden, tossed below the required height, or tossed below table level. The ball touching clothing, the body, or the net. Score disputes. Each player has a limited number of challenges, usually one or two, and loses one if the challenge fails.

What stands out is the standard written into the regulations. To overturn a decision, the panel must determine that the umpire made a "clear and obvious error". The phrase is borrowed from other sports, where it is no less contested. It sounds like a technical threshold. It is not.

Tactics are what people draw on a whiteboard during a timeout. Data is what they draw on the table surface.

Data: what the notebook recorded

Since the start of the season, I have logged every video review in matches I watched live or on full recordings. The sample covers 214 matches at WTT-series level, continental qualifying events, and a number of Chinese domestic league matches. In total, 1,087 reviews were recorded across four data fields: the moment within the game, the type of situation, the processing time, and the announced outcome.

Situations fall into four broad categories. Edge-of-table or side-edge contact accounts for 41 percent. Service-related reviews — hidden ball, toss, hand position — account for 34 percent. Ball touching the body or clothing accounts for 9 percent. The rest are score disputes, net contact and rare situations.

Overturn rates tell the story far better than the aggregate number. Across the full sample, 36 percent of reviews led to a decision being reversed. Broken down by category, the gaps open quickly. Edge contact: 52 percent. Service: 19 percent. Body contact: 44 percent. Everything else: 23 percent.

The gap between 52 and 19 does not measure umpiring quality. It measures the difficulty of the question. An edge ball is a binary question: it touched or it did not, and a single frame from the right angle can answer it. A service is a continuous question: what percentage of the ball must be hidden before it counts as hidden? At what height must the forearm sit before it obstructs the receiver's view? Every umpire has a different threshold. The regulations do not quantify that threshold, because quantifying it is extremely hard.

Overturn rates reflect the structure of the question, not the competence of the person answering it. That is the central finding of the whole season, and it renders every umpire ranking I have ever seen meaningless.

There is an even lower layer of data that I struggled to capture. The contact between hand and ball during a toss lasts a few hundredths of a second. At standard broadcast frame rates, that interval spans only two or three frames, and the frame containing the release is usually blurred by motion. Many tournaments run their review system on the broadcast feed itself, not on a dedicated high-speed camera rig at every table. The hardest question is answered with the weakest tool.

I spent four evenings checking this by counting frames manually on recordings. Of 63 service situations sent to review, I could identify a clear contact frame in 21 cases. In the other 42, the contact frame was blurred, or the ball had already left the hand before the first frame of the replay clip. I am not concluding that the decisions in those 42 cases were wrong. I am concluding that they could not be verified using the very footage the panel used.

And there is one more layer, the layer that made me abandon any attempt to rank umpires. My tracking sheet records whether a decision was overturned, not whether it was correct. Both directions are biased. A correct decision can be overturned because the angle was poor. An incorrect decision can stand because the image was ambiguous. There is no "truth" column in my sheet, and I suspect there is none in anyone else's.

My tracking model has no heart, which is precisely why it can never be hurt — and also why it can never touch the only thing worth measuring.

I still remember an evening at the Longgang district arena in Shenzhen, sitting next to a retired international umpire. He rewatched a rally with me, rewinding four times, then said something I copied verbatim into my notebook: "You ask me whether the ball touched the edge. I can answer that. You ask me whether his forearm hid the ball. I can only tell you that from my angle it did, and from the angle of the person on the other side of the table it did not."

That is an answer about geometry, not about morality.

In 2026, a press conference door closed in my face. Today I read that moment back through data, and I realise it was also an answer about geometry: I was standing in the wrong place, and no camera frame recorded the question I had asked.

Tempo: what the scoreboard does not record

A video review lasted an average of two minutes and forty seconds in my sample, measured from the moment the player raised a hand to the moment the chief referee announced the outcome. That figure varies widely by tournament: at events with larger panels, the average rose to three minutes and twenty seconds.

I separated matches with a final margin of four points or fewer and compared them with the rest. In the close group, the average number of reviews per match was 8.7; in the rest, 4.3. Anyone who watches table tennis could have guessed this. Pressure increases challenge volume. There is no surprise here.

What made me stop was what happened afterwards. I measured the average rally length across the three points immediately following a review and compared it with the average across the three points immediately before it in the same game. Across the sample, rally length fell from 5.4 strokes to 3.9. The share of points ending within the first three strokes rose from 38 percent to 51 percent.

There are at least two explanations, and I do not have enough data to choose between them. First: the break cools the hand, breaks the feel, so rallies shorten. Second: the player who won the review accelerates to ride the psychological momentum, while the player who lost tries to end points early and escape the noise.

A third explanation is less attractive but more honest: both players are hitting in a state of disrupted attention, and table tennis at this level punishes disrupted attention faster than it punishes any technical error.

Players leave the table, spectators leave the stands, but data never leaves the game.

One player, one story, and one variable nobody logs

Mid-season, a female player ranked inside the world's top twenty returned from a left wrist injury that had kept her out for nearly four months. The team's statement said she was "ready physically and mentally". She came back at a small WTT event, not on a big stage.

In her first four matches after returning, she was called for eleven service faults. Her own average the previous season was 1.8 per match. The number multiplied several times over, and online commentary immediately shifted to the theme that umpires were watching her more closely than others.

I rewatched all four matches, frame by frame. Her service motion had changed. In the old motion, her left forearm rose high and outward before the toss, keeping the contact zone completely open. After the injury, the range of forearm lift was roughly fifteen degrees shorter, and in three of the four matches the forearm stopped at chest height, creating a small blind zone behind the ball during the contact window.

The umpire was not wrong. She was not cheating. A wrist injury had changed the geometry of a movement, and the new geometry violated a rule written to deter deliberate behaviour.

We are reading a biomechanical problem in the language of a moral problem, and then wondering why the argument goes nowhere. My tracking sheet captured the spike in fault calls. It could not capture the reason, because the reason sits in a wrist joint no camera is pointed at.

Her return timeline, to put it bluntly, was written more by the communications department than by the doctor. The phrase "waiting until the weekend" appeared twice that month, and both times it pushed the return back by another week. A wrist injury in table tennis does not need a long rest to heal; it needs a long rest to return to the same range of motion. That is a kind of time no press release quantifies, and no tracking sheet measures.

The counterintuitive angle: consequence is not cause

I presented part of this sample at a small talk in Shenzhen, and the first question from the audience was whether umpires had got worse this season. The overturn rate rose, so the question seemed reasonable.

It is not reasonable.

The overturn rate is a fraction, and its numerator depends on the players' decision about whether to challenge at all. A season can show a higher overturn rate because umpires erred more, or because players challenged more, or because they challenged more cleverly, or because a new cohort of players emerged with service techniques that push more situations into the grey zone.

This season alone, I have seen evidence for at least two of those four possibilities. Challenges per match rose 22 percent year on year within the tournament tier I track. And the success rate among players under 23 was markedly higher than among players over 28, even though the younger group challenged less in absolute terms.

That is a signature of learning, not of umpiring quality. Younger players watch more footage, and they know exactly which situations carry a high probability of being overturned.

The "clear and obvious error" standard cannot function as a technical threshold, because it has no unit of measurement. It functions as a negotiation. Three people sit in front of a monitor, watch the same clip, and bring three different bodies of experience about what counts as "obvious". The outcome of that negotiation depends on the composition of the panel, on the pressure of the moment, and on how many cameras the arena happens to have in the right place.

I am not saying panels act arbitrarily. I am saying the wording in the regulations conceals a fact: the final decision is made by people, not by frames.

And if my numbers are wrong? That possibility is not small. My four data fields do not include panel composition, camera count or placement, or whether a player had already exhausted their challenges. A sample of 214 matches is enough to reveal structure, not enough to close the argument. If someone has a fuller dataset and reaches a different conclusion, I will be the first to rewrite this piece.

Looking to the next round

Three signals are worth watching.

Scope of review. If the system expands to more complex situations — say, disputes over whether the ball touched the table twice during a fast rally — overturn rates will fall rather than rise, because the number of questions video cannot answer will grow.

Publication of review data. No tournament currently publishes aggregate review figures, even anonymously. If that changes, the analytics field will finally have a basis for comparing across events and across seasons.

Injury-layer data. A player returning with a changed movement will generate a wave of technical faults across the first four to six matches. That signal is predictable, and it is being misread every single time.

There is a larger question beneath all of this. Table tennis has built a review system on the assumption that every disputed situation has one correct answer, and that the job of technology is to find it. After a season of tracking, I believe that assumption holds for roughly sixty percent of situations, and fails for the rest.

For the rest, the most honest answer an umpire can give is: I do not know. No sport has yet written a rule that allows anyone to say that.