The Fifth Set and Vietnam's 17.2-Point Gap in Table Tennis

The fifth set, score 9-9. A Vietnamese player steps up to serve, and in my ledger that is the worst cell of the entire match. I tracked 11 matches at the national championship and the last two SEA Games, logging every point across four variables: server, serve type, rally length, point winner. Across the first four sets, Vietnamese players won 61.4% of the points in their own service sequences. From the ninth point of a deciding set onward, that number falls to 44.2%. A drop of 17.2 percentage points. Singaporean and Thai players in the same sample drop 6.8 points. That 10.4-point gap is not a serving-technique problem. It sits somewhere else, and it took me nearly three seasons to pin it down.

Vietnamese table tennis is not short of good players. Nguyễn Anh Tú, Đinh Quang Linh, Trần Tuấn Quỳnh and Nguyễn Đức Tuân on the men's side; Mai Hoàng Mỹ Trang and Nguyễn Khoa Diệu Khánh on the women's — names that have kept the national team inside the regional medal group for years. Their problem was never a shortage of shots. It is what happens after the fourth shot.

I began building my own database in 2026, after realising that Vietnamese table tennis has almost no public point-level data. The world federation publishes results, not point structure. Television shows matches, not rally sequences. To find out where a player loses points, someone has to sit and count. I sat and counted. Each match is a sheet of paper split into four columns, and by now I have 11 matches clean enough for analysis — clean meaning there is video, there is a point-by-point scoreline, and not a single point is missing. The sample covers matches at SEA Games 31, held in Hanoi in May 2026, and SEA Games 32 in Phnom Penh in May 2026, plus national championship finals and semi-finals.

The first result that jumped out was rally length. Vietnamese players win 48.9% of points in rallies of four touches or fewer. From seven touches upward, that rate falls to 41.3%. In the Singapore and Thailand control group, the same two figures are 47.6% and 46.1% — almost flat. The difference between a player who can hit a ball and a player who can win a match sits exactly in the gap between those two numbers. The longer the rally, the faster decision quality falls — and it falls faster than the opponent's.

The rhythm of the set traces a similar curve. I split each set into three blocks: 0-0 to 5-5, 6-6 to 9-9, and 10-10 onward. In the first block, the Vietnamese group's point-win rate is 52.1%. The middle block falls to 49.4%. The final block falls to 43.8%. The control group goes 51.2%, then 49.8%, then 48.9%. The opponent's curve is nearly horizontal; ours slopes down exactly in the block where one point buys a set.

What pushed me to write this piece is timeout usage. Across 11 matches, Vietnamese players called nine timeouts. Seven of those nine came after the opponent had built a lead of at least three points — that is, once the match had already tilted. Only two came with the score still level. In the control group, six of eight timeouts were called with the gap at one point or less, usually right after the opponent had won a long rally. A timeout is not a life raft; it is a change of rhythm, and we are using it as a life raft.

This is where the story is easy to misread. The familiar reading is “no nerve”, and I have heard it often enough to know it leads nowhere. My data does not support it. If the problem were purely psychological, the drop would appear evenly across every high-pressure point, including 10-10 after a short rally. Instead, the Vietnamese group's drop is concentrated almost entirely in points that follow a long rally, and it disappears in points that follow a short one. In other words: the player stays calm, the legs simply stop arriving on time.

I tried to verify it another way. In six matches with sufficiently long video, I measured each player's lateral movement per set by counting how many times they crossed the vertical lines along the table. Sets one and two were nearly identical. Set three dropped 7%. Set four dropped 14%. Set five dropped 23%. That is a fitness curve, not a nerve curve. And it lines up precisely with where the point-win rate collapses.

A fitness hole never shows up in the ranking table; it only surfaces in the fifth set. The ranking records the match result. It does not record that the result was settled in the fifth set, after the legs had tired and the decisions had slowed.

Stopping at fitness would leave out the second half of the problem. Fitness explains why the legs slow. It does not explain why slow legs cost points inside the service sequence — something that, in theory, depends on the arm and on tactics. The answer sits in the structure of the domestic calendar.

The national championship currently runs best-of-five in most events, and most matches end in three or four sets. That means a player competing through an entire season may only touch a 9-9 deciding-set situation a handful of times. When a situation rarely appears, it is not trained; when it is not trained, it is handled on instinct; and instinct carries no data. Every player has a weak joint; my job is to find it before the opponent does — and this particular joint is not in anyone's arm, it is in the schedule.

There is an objection I have to raise before anyone else does: 11 matches is a small sample. True. With 11 matches I cannot assert anything with certainty; I can only speak in probabilities. But a small sample does not erase the direction of a curve. All three independent indicators — rally length, score block, lateral movement — point the same way. When three different measurements return the same answer, the odds that all three are wrong are lower than the odds that one of them is right.

There is one more variable I once overlooked: the crowd. In matches with a home crowd, the Vietnamese group's point-win rate at 10-10 is 47.1%; without a crowd, it is 42.6%. A 4.5-point gap — small, but pointing the same direction as what I saw in football during the empty-stadium period. The crowd is a variable in the equation, and noise is only the audible part of it. Remove them from the equation and a correct conclusion can still be wrong about its size.

So what should be measured in the next cycle? Not wins. Wins are outcomes, and outcomes are visible to everyone. What needs measuring is the point-win rate from 9-9 onward, tracked per player, alongside lateral movement in the fifth set. Those two together will forecast a player's ability to reach a final better than any ranking table.

And there is one small technical change I consider worth trying: move some domestic events to best-of-seven. Not to make matches longer for entertainment, but to make the 9-9 deciding-set situation appear twice as often — and become something trained rather than something endured. A season is a long chain, but people only remember the last three matches. Those last three are settled in the sets we currently do not practise enough.

I do not believe in form; I believe in form data, and the two rarely agree — least of all in the fifth set. Next season, when the scoreboard opens again at 9-9, I will still be sitting there with a sheet of paper split into four columns, and the only question I want answered is whether that curve has started to flatten — or whether we will simply give it another name.