Table Tennis, Empty Data, and the Discipline of Not Guessing

In July 2026, in Nha Trang, I opened a fourteen-page data package on an international table tennis stage. The cover had a logo, the inner pages had tables, the formatting met publishing standards. I read from the first line to the last and wrote exactly one sentence in my notebook: not a single usable information point. Set scores were there, but without point distribution by phase. Player names were there, but without accumulated match counts, without ranking-point expiry dates, without a single indicator on service performance. The commentary section was full of adjectives: "character", "decisiveness", "class". Meanwhile, the market line kept ticking upward hour by hour, and bettors kept making decisions.

I have met this scene many times in nearly a decade of betting analysis. It forces me to restate a principle that has followed me since 2026, when I built my own xG spreadsheet for a round of V-League fixtures: a document without information points is not a document, it is a form of presentation. In table tennis the principle is stricter still, because this is a sport Vietnamese fans follow in huge numbers while public data stays thin. That gap is where bad models are born, grow up, and get passed around as truth.

Table tennis reaches Vietnamese audiences through three main channels: score feeds, short highlight videos, and commentary pieces that are translated or freely adapted. Only the first channel carries primary data. The other two carry interpretation, and interpretation cannot be verified. In football, a viewer in Nha Trang can still look up shots, pass completion, distance covered. In table tennis, even events under the WTT system publish only a fraction: game scores, sometimes point scores per game, rarely point-by-point detail. To get data, I have to sit down with video and count by hand.

The history of the rules explains why table tennis data reads the way it does. In 2026, the ITTF increased ball diameter from 38mm to 40mm. In 2026, games were shortened from 21 points to 11, and the service rotation from 5 points to 2. In the early 2000s, the ban on hiding the ball at service was tightened. In 2026, plastic balls replaced celluloid. Those four changes together sharply reduced direct service winners, increased extended rallies, and turned the ability to survive a rally into a measurable capacity. The sport was redesigned so that data would matter more — but the data infrastructure did not keep up.

From 2026, WTT has operated a tiered event system in place of the ITTF World Tour, with tiers running from Grand Smash down to Contender and Feeder. Each tier carries a different point value, and the ITTF ranking counts a player's best results over the last twelve months, rolling weekly. That mechanism has a consequence few followers notice: a player's ranking does not reflect current form, it reflects that player's competition schedule over the past twelve months. Points are not lost because you played badly. Points are lost because you did not play.

Domestically, Vietnamese table tennis has a broad grassroots base, a national championship system, youth age groups, and a handful of names that have drawn attention on the regional stage. But detailed statistics at domestic event level are almost entirely absent in public. No point-by-point data, no rally-length distribution, no injury records. That means anyone speaking about Vietnamese table tennis in numbers is speaking about data they collected themselves, and readers are entitled to ask: collected how, over how long, and is the method written down.

If I had to build a working analytical framework for table tennis, I would start from the structure of points, not from the score. Each game holds a few dozen discrete points, and an 11-9 game carries entirely different information from an 11-3 game. The final scoreline hides the whole shape inside. So I always split data across three game phases: from 0 to 4 points, from 5 to 8, and from 9 onward. Win rate in the closing phase is the closest thing to the word "character" that commentary pieces use impressionistically. The only difference is that this indicator can be verified, and impressions cannot.

The first three shots are where the match is written. Serve, receive, third ball. From video data across WTT matches that I counted myself, the service-point win rate of leading players usually sits between 55 and 60 percent, and that rate varies sharply by spin type. A player might win 70 percent of points serving sidespin and only around 45 percent serving topspin. If a piece of analysis does not separate by spin type, it is describing someone other than the person walking to the table.

Rally-length distribution is the cheapest and most useful indicator in this sport. I count it by hand from video, exactly the way I once counted Croatia's passes in 2026. Croatia 2026 taught me that a pass under pressure is a statement, not merely a technique; a receive under pressure is the same. A player who wins most points in rallies of 1 to 3 shots lives on serve and third ball. A player who wins in rallies of 5 shots or more lives on durability. Those two players require two different counter-plans, and a model that does not separate these groups will produce errors of the same repeating kind.

Equipment is the most underrated variable. Rubber thickness, pips type, glue, and even the racket inspection process all bear directly on the data. Changing rubbers mid-season is a technical change that needs adaptation time, like a striker changing boots. In my records, the three-to-five weeks after a player switches rubbers typically produce large swings in third-ball point win rate. That is verifiable information, not speculation.

For each player I track three data groups. The first is ranking and points structure: current position, points expiring within sixty days, events that must be played to hold position. The second is head-to-head, split by event, by table and ball conditions, and limited to the last two years. The third is performance in decisive matches, especially seventh games in best-of-seven formats. These three groups answer three different questions, and blending them is the fastest route to a meaningless conclusion.

The event system determines the value of every result. A win at a Grand Smash and a win at a Feeder differ many times over in points, and they differ in opponent quality too. When I assess a player, I always ask which tier that good run happened at. The same question applies to entry choices: skipping a low-tier event to concentrate on a high-tier one is a strategic decision, and it leaves a clear trace in ranking data a few months later.

On the international scene there is a relatively clear axis. China holds most of the leading positions, with a roster depth no other association currently matches. The challenger group includes Japan, South Korea, Germany, Sweden, France and Brazil, each strong in a different player type. When analysing a non-Chinese player, I always check one variable with high explanatory power: win rate against Chinese players over the last two years, split by right-handed and left-handed. Names such as Ma Long, Fan Zhendong, Tomokazu Harimoto, Truls Möregårdh or Hugo Calderano only carry statistical meaning when read against that indicator.

Rules and governance are the least-discussed layer and the one that directly affects predictions. Racket inspection, rubber thickness limits, service rules, interval timing between games, and disciplinary rulings can all swing a match. A player cautioned over a service motion must modify it, and that player's service-point win rate typically drops over the next two or three matches. Aggregate data cannot capture this kind of information. Only match-by-match note-taking can.

Coaching staff and development systems are the hardest layer to quantify and impossible to skip. At national-team level the central question is conversion efficiency: how many junior players break into the senior squad and stay there for three seasons. At club level the question is stability. A head coach changed mid-cycle usually drags a training-structure change with it, and individual performance data moves in the weeks that follow.

The risk surface has several overlapping layers. Injury risk attaches directly to schedule density: the 11-point format makes each game shorter, but the number of matches per week in the WTT system is higher. Equipment risk comes from adaptation time. Tactical-decoding risk comes from video, since every service pattern gets reviewed and countered within a few events. Data-source risk is rarely mentioned, even though it sits behind every other layer.

Media narrative and market expectation run to their own rhythm, and that rhythm is usually faster than the rhythm of data. A young player winning a few matches at a small event gets described as a phenomenon. An older player losing once gets described as declining. Both descriptions lack sample. With the data I collect, roughly ten matches are needed before a trend can be discussed, and even then, opponent-quality variance remains a bigger noise source than form variance.

Industry transmission is the layer the Vietnamese market follows least and the one that affects viewers most. Equipment change at the professional level filters down to grassroots play within two to three years. The rise of a player with an unusual style changes how children are coached. An event staged successfully in one city pulls demand for tables, balls and classes. These chains never show on the scoreboard, yet they shape the scoreboard of five years from now.

The most counterintuitive thing I have learned in table tennis is that the biggest risk is not a wrong prediction. Wrong predictions are normal in a sport where each point lasts a few seconds. The real risk is a wrong model. I fear a wrong model more than a wrong call, because it is wrong systematically. A wrong call costs me once. A wrong model costs me ten years, and worse, it teaches me lessons that do not exist.

So the most serious risk layer in table tennis sits in the data source, not in the sport itself. No point-by-point data, no rally distribution, no public injury history. When the source is empty, every conclusion becomes speculation dressed as analysis. And the pressure to fill the gap is the strongest pressure in this profession, stronger even than the pressure to be right. A writer wants an ending, so the writer manufactures data. I have set myself one rule: when the source is empty, the only correct conclusion is to state that the source is empty.

Another trap is correlation. In my data, players who win more long rallies tend to win matches. That easily leads to the conclusion that winning requires extending rallies. In reality, most long rallies occur in matches where the stronger player is ahead and is deliberately playing safe to protect the position. The cause lies in match state, not in rally length. A model reading this chain backwards will advise a player to do precisely the opposite of what is needed.

Finally, there is one thing about Vietnamese audiences that anyone working with data must accept. Table tennis viewers follow flags, stories, faces. They do not come for indicators. If I open with a table, I lose them by the third line. If I tell a match moment first and then put the indicator on the table, they stay. Data does not forgive emotion. And that is why I converted. But for the same reason, data has to be told in human language, or it is just a spreadsheet defending itself.

The signal for the next round is not which player is in form. It sits in two other places. The first is the calendar. With a rolling twelve-month points mechanism, the next two months are a period when old points expire en masse for many players, and the rankings will shift according to schedule before the first ball is even tossed. The second is data. If international event systems begin publishing point-by-point detail, the entire table tennis analysis industry will change within two seasons, and the fastest to adapt will be those already used to sitting down with video and counting by hand.

I still keep the habit of opening every data package with a single question: how many information points are in here. If the answer is none, I write exactly that and add nothing. A goal is only the conclusion. xG is the testimony. And in table tennis, the testimony lies in the rallies nobody bothers to count.