Vietnamese Table Tennis and the Data Tribunal: Why the Closest Sets Hide the Widest Gap

One evening in May, I stayed alone in the analysis room after the screen had gone dark. On the desk there was only a notebook and a spreadsheeet left open. The match I had just watched ended with a score that silenced the stands: a Vietnamese player lost narrowly after seven sets, including two sets won 11-9 and one set lost 9-11. The ordinary spectator left the arena with a sense of "regret", with the belief that just a bit of luck, a more precise wrist flick, would have changed the result. And I, once I had downloaded the raw point-by-point data of the entire match, saw a story pointing in the opposite direction.

Seven sets, twenty-one points won on narrow margins, and a gap so large it did not lie in the score. It lay in a column no one in the stands could see: the opponent's point-win rate on serve. When I placed that column beside the Vietnamese player's corresponding column, the picture became sharp to the point of discomfort. The closeness of the score is an illusion created by a probability equation, not an indicator of equal ability. That was the first line I wrote in my notebook. Numbers never lie; only the reading of them is wrong.

Before going into the core, I need to set the context and method. Table tennis is analyzed far less through data than football, basketball or tennis. There is no xG for table tennis, no globally standardized statistics system like Opta or StatsBomb applied to this sport across every tournament. That means the analyst must build a set of metrics from the rawest data: every point, every serve, every stroke. I work this in three layers. The first is raw point data, which anyone with a results sheet can collect. The second is action data: serve type, placement, receive direction, and who struck first in the rally. The third, and hardest, is rally-sequence data: the length of each exchange and who won the long rallies.

Based on my experience following matches across regional and international seasons, I have found that Southeast Asian table tennis in general, and Vietnamese table tennis in particular, has a very clear data signature: results in short rallies are better, but results in long rallies drop sharply. This is something the naked eye never catches, because a three-beat rally and a seven-beat rally, once the point has fallen to the opponent, look exactly the same on the scoreboard: one point for them. Only data can separate those two kinds of points.

Let me start from the broader context. World table tennis is in a structural transition. The World Table Tennis (WTT) system was designed to increase the number of matches and the density of competition. In theory, greater density produces more data, more matchup samples, and more chances for players outside the dominant group to accumulate points. But the macro-level data shows the opposite picture: as the number of matches rises, the gap between the leading group and the chasing group does not narrow; it is magnified. The reason is not the number of matches. The reason is the quality of information teams extract from those matches.

The current dominance of Chinese table tennis, with names like Ma Long, Fan Zhendong and Wang Chuqin, is not only about technique. It is about data infrastructure. Each of their matches is filmed, analyzed, and converted into training data. A Vietnamese player in the same period may play an equivalent number of matches, but the amount of information extracted from each match is many times smaller. The result is a compounding effect: after three years, the accumulated gap in "data capital" becomes a gap in performance, even if the hours spent training at the table may be similar.

Regional context matters too. In Southeast Asia, table tennis from Vietnam, Thailand, Singapore and Malaysia compete closely at the SEA Games. At the Asian and world level, the gap is much larger, but precisely for that reason the regional tournaments become an important laboratory. A player can only improve their data when that data exists. And data only exists when someone stays behind after the match, dissects every point, labels every action, and turns moving images into columns of numbers.

At this point I must be clear about my own limits. Not everything in table tennis is measurable. Dynamic psychological states, the feel for the ball on a given day, crowd pressure, and accumulated fatigue after many matches are all extremely hard to encode. But hard to encode does not mean impossible. Pressure can be approximated by error rate at deciding points. Fatigue can be approximated by average rally length and win rate in the second half of a set. Feel for the ball can be approximated by the direct-point rate on serve. No variable is fiction as long as we define it by a specific measurement and keep that definition fixed across matches.

With that approach, I built a core metric set for table tennis with five groups. The first is serve efficiency: the direct-point rate on serve and the point-win rate on the third ball. The second is receive efficiency: the point-win rate when the opponent serves and the rate of forcing the rally into longer exchanges. The third is rally structure: average rally length and win rate by length threshold. The fourth is movement efficiency: estimated steps per set and the rate of maintaining balance after the first step. The fifth is clutch efficiency: points from 9-9 onward in each set and deciding points in the seventh set.

When I applied this metric set to the regional matches I followed, a recurring pattern appeared so often that I had to re-check the data several times. In the first and second groups, the leading Vietnamese player's metrics were no worse than those of Thai or Singaporean opponents. In some matches, their direct-point rate on serve was even higher. This explains why spectators always feel it is "evenly matched". But moving to the third and fourth groups, the gap opens clearly. In rallies of seven beats or more, the win rate of the Vietnamese group falls sharply. And in the fifth group, efficiency at points from 9-9 onward drops yet another tier.

Every tactic is only a hypothesis until data delivers its verdict. And the data here delivers a clear verdict: the problem lies in the ability to sustain structure in long rallies and in the ability to hold that structure under the pressure of deciding points. These are two different problems, and they are often wrongly merged under the label "weak mentality".

I want to separate those two problems with a concrete example from the data table. Imagine a player with a win rate of 58% in short rallies (one to three beats), 50% in medium rallies (four to six beats), and 38% in long rallies (seven beats or more). This is a very typical pattern for a player with good early attacking technique but a weak foundation in extended rallies. Now place beside that an opponent with the opposite pattern: 44% in short rallies, 52% in medium, and 61% in long. When these two meet, the result is not a balanced match. The result is a match in which the first player must finish the point as early as possible, because the longer the rally runs, the lower his win probability.

This is the mechanism behind what I call the "illusion of closeness". A player A beating a player B 11-9 in a set does not mean they are equal. It means that in that set, the short rallies fell to A more often, and A capitalized. In the next set, if B adjusts to extend rallies, the score may still be 11-9, but this time in B's favor. And by the seventh set, when both know each other's weaknesses, B only needs patience to stretch the rallies and win. The spectator sees seven balanced sets. The data sees a player forced to win by the hardest route in every set.

Let me go deeper into the third metric group, rally structure. In modern table tennis, the average rally length at the top level fluctuates within a relatively narrow band. But it is precisely within that narrow band that differentiation becomes clear. A player with an average rally length of 3.8 beats and a 55% long-rally win rate is a player with a durable foundation. A player with an average length of 3.2 beats and a 40% long-rally win rate is a player who lives by speed, and who will die by speed when the opponent denies him speed.

The interesting thing is that both players can have the same overall win rate. The same performance, two entirely different profiles. And in the transfer market, in selection work, in building a national team, it is precisely the difference in this profile that must be read, not the overall performance. A team made up entirely of speed-based players will be very strong against weaker opponents but very fragile against opponents who know how to extend rallies. A team with a blend of speed players and durable players will be far more flexible.

This is where I must address one of the most widespread beliefs in Vietnamese table tennis, and the data truth behind it. The belief is: "We lack luck at deciding points." I have heard this hundreds of times. And I will say it plainly: luck is the name we give to a process we have not measured. When I encoded deciding points as a specific variable, that is, the win rate at points from 9-9 onward, I found a pattern that was far from random. Players with a low long-rally win rate also had a low deciding-point win rate. The correlation was very tight. This means the problem at deciding points is not an independent psychological problem, but a consequence of a structural technical problem: when pressure rises, a player tends to return to his safest habit, and if that safest habit is to finish points early, then at deciding points he will try to finish early, and the win probability will fall exactly to his short-rally win rate.

I must be careful here not to turn correlation into causation. The fact that a low long-rally win rate accompanies a low deciding-point win rate does not automatically prove that long rallies cause the deciding-point problem. There is at least one alternative hypothesis: perhaps both are consequences of a common third cause, such as physical foundation or the quality of footwork in the second half of the match. Or perhaps both are consequences of training quality in pressure situations. That is why I always keep a "rebuttal card" beside every conclusion: an uncontrolled variable that could overturn the story. In this case, that variable is accumulated minutes played in the tournament, which I cannot yet fully measure.

Let me move to the fourth metric group, movement efficiency. This is the most overlooked group in table tennis, perhaps because it is hard to measure. In football, distance covered is a mainstream metric. In tennis, distance moved per point has long been measured. In table tennis, people rarely talk about it, even though table tennis is a sport that demands extremely fast repositioning over a small area. A player shifts from attack to defense and back in less than a tenth of a second.

When I estimated steps per set from video, I found a notable pattern. Leading regional players had a higher number of steps per set, but not linearly higher. What matters more than the number of steps is the rate of maintaining balance after the first step. Strong players place their first step so well that the second and third steps are shorter. Weak players' first step is not good enough, so they must compensate with many small steps, burning energy and losing preparation time for the stroke.

This data connects directly to the third metric group. When the movement foundation is weak, long rallies become more energy-expensive, giving the player an incentive to finish early. But finishing early means accepting higher risk, and higher risk at deciding points is a recipe for losing. The loop closes here: weak movement leads to short rallies, short rallies lead to high risk, high risk leads to failure at deciding points, and that failure reinforces the belief that the problem is psychological.

I want to devote a section to equipment, because in table tennis equipment is not a minor detail. Rubber and blade form a system in which every change brings an adaptation period. When a player switches from a high-spin rubber to a more control-oriented rubber, he typically needs time to adjust his contact point. During that period, his data will temporarily worsen in the serve metric group and the short-rally group, while the long-rally group may improve. If the analyst looks only at overall performance during that period, he will wrongly conclude that the player is declining, when in fact he is in the process of transitioning to a better profile.

Based on my experience following matches, very few table tennis teams in Southeast Asia have a process for measuring equipment adaptation periods. This leads to an interesting paradox: equipment changes that are beneficial in the long run are often judged as failures in the short run, because no column of numbers can separate "adaptation period" from "decline". Once again, numbers never lie, but when numbers are missing, belief fills the gap.

Now I want to move to the counterintuitive part, the part I believe is most important in this entire analysis. The popular belief is that to close the gap with top table tennis, we need to strengthen attack, find sharper servers, stronger wrist flicks. My data suggests the opposite.

In modern table tennis, what separates players is not the ability to create an advantage, but the ability to survive when you have no advantage. Read that carefully. All players at the regional level can create an advantage at some points. What distinguishes winners from losers is who can sustain structure when pushed into a disadvantage. Attack is the ability to create advantage; that is the part everyone trains. But surviving when advantage is lost, restructuring a rally when you have been pushed back, and turning a defensive situation into a neutral rally to start again, that is the rare skill.

Defense is a poem written in data. In table tennis, defense is not measured properly. No one counts how many times you drag a rally from disadvantage to neutral. No one counts how many times you force the opponent to play one more beat and make an error. But those moments are exactly what create the long-rally win rate, and as shown above, the long-rally win rate correlates tightly with the deciding-point win rate.

This is where I want to push the paradox one step further. If defense matters so much, why is it undervalued in training? The answer lies in the structure of youth competition systems. Youth tournaments tend to reward attacking play because at that age technical gaps are still large and a strong attack can overwhelm weaker opponents. This creates a training loop: young players learn that attack is the path to winning, so they train attack, so when they reach a higher level, where everyone defends well, they lack the foundation to survive. The truth is that what wins at youth level is not what wins at senior level.

I have seen this pattern more than once. A young player can win a youth title with an extremely high short-rally win rate, then struggle greatly when stepping onto the adult stage, where opponents do not make easy errors. And when that happens, people talk about "the young player's psychological crisis", when the real problem is a skewed technical profile built over years.

There is one more layer to this paradox, related to the psychological factor I mentioned but have not fully explored. Psychology in table tennis is not a variable standing outside the data. It is a variable to be encoded, like any other. Error rate at deciding points, time between points, the number of tempo changes when trailing, and the tendency to choose the safe option when trailing can all be measured. When I measure them, I find that players who tend to switch to the safe option when trailing lose more, even when the safe option theoretically has a lower error probability. The reason is that the safe option often hands control of the rally to the opponent, and at the highest level of table tennis, handing over control is equivalent to handing over the point.

This leads to a concrete proposal that I believe is valuable to any training system: instead of measuring overall win rate, measure win rate in specific states. Win rate when leading, win rate when trailing, win rate in short rallies, win rate in long rallies, win rate at deciding points. A player with the same overall win rate can have two entirely different portraits when viewed through these states, and that portrait is what determines their future.

Let me synthesize the chain of evidence before moving to the conclusion. One, the short-rally win rate of Vietnamese players is at a regionally competitive level. Two, their long-rally win rate is significantly lower. Three, the long-rally win rate correlates tightly with the deciding-point win rate. Four, the movement foundation explains part of the long-rally problem. Five, equipment adaptation periods create temporary data fluctuations that are often misread. Six, the youth training system rewards attack, creating a skewed technical profile in the long run. All six points are not isolated observations. They are a system.

And this is where I must speak about the nature of that system. None of the points above is a problem of an individual player, nor of a specific tournament. It is a problem of a process. Vietnamese table tennis lacks a data infrastructure. Data does not save a season, but it points out exactly where the season dies. And if we do not know where the season dies, we cannot save it next time.

I want to use the final space to talk about the hardest thing: the difference between having data and using data. Many teams have video, score sheets, notes. But data only has value when it is structured, labeled, and compared correctly. An unlabeled video is a memory. An unlabeled score sheet is a history. Only when you add the question "why was this point won" do you begin to have data. The data ocean is not for those afraid to get wet.

The upcoming transfer window and competition cycle are an opportunity to change this. Not by buying better players, but by building the ability to read players better. A player whose technical profile fits the system will bring more value than a player with higher overall performance but a skewed profile. This is why I say the boldest decision in a transfer window is not paying the highest price, but refusing to pay for the wrong profile.

Recognition comes late, but data is always on time. On a personal level, I still keep the habit of staying behind after every match, even hours after it has ended, just to label a rally my eyes missed. Because after many years, I have learned that what the stands call "regret" is often a column of numbers that has not yet been seen.

I do not believe in fairy tales. I believe in the chain of events that leads to a point. And in table tennis, that chain of events always begins in a place no one bothers to look: at the moment the foot lands on the floor after the previous stroke. That is the influence of the first step, and it determines everything that follows.

There is a question I want to leave for anyone involved in training or selection: if you could keep only one single column of numbers to evaluate a player, which column would you keep? If your answer is "win rate", you are looking at the results of the past. If your answer is "long-rally win rate at deciding points", you are looking at the future. Between those two answers lies the entire gap that Vietnamese table tennis needs to cross.

And this is the signal I will track in the next round of competition. I will track the long-rally win rate (seven beats or more) of the country's leading players in matches against regional opponents. If that column improves, I will believe in a near future. If it stays flat, then any progress in other columns is just a new coat of paint on an old wall. Data will deliver its verdict, and this time, I will read every line carefully.

Finally, I want to speak about the broader meaning of this story. Table tennis is a sport that demands a combination of technique, physicality, tactics and mentality. Data analysis does not diminish its beauty. On the contrary, data makes that beauty clearer. When you understand why a long rally was won, you understand more deeply the skill that created it. When you understand why a deciding point was lost, you understand more clearly the pressure the player had to bear. Data does not destroy the legend of a wrist flick. It retells the story behind that flick in precise language.

And that is why I do this work. Not to reduce sport to dry numbers, but to restore the truth behind the moments that emotion has obscured. Every player walks onto the court with a story, and data is the way that story gets told correctly. Vietnamese table tennis has stories that deserve to be told correctly. My task, and that of anyone who works with numbers, is to build the infrastructure so those stories are not lost.

Because after all, the worst thing is not losing a match. The worst thing is not understanding why you lost, and therefore never winning.