The Empty Data Column and the Double-Verification Standard in Table Tennis Writing

That night, after the final round of group play at a national table tennis event, I opened my notes file and found the column "points won on serve" completely blank. Forty serves across three games, and my recorder had stopped midway through the second game because the organisers moved the match to the far row of tables. There were two ways to handle it. The first: fill in a plausible-looking number — 62%, say — and keep writing, since nobody reading a quick news piece would check. The second: leave the cell empty, mark it "insufficient data", and accept that the piece would look less dense than someone else's. I chose the second. The next morning my editor called to ask why an analysis piece had holes in it. My answer was the one I have used across nine years of podcasting: a fabricated number wearing a high-confidence label is worse than no number at all.

Table tennis analysis in Vietnam is in an odd phase. Audiences are growing, the domestic calendar is steady, and players such as Nguyễn Anh Tú, Đinh Quang Linh, Mai Hoàng Mỹ Trang and Trần Tuấn Quỳnh now carry enough individual following to pull tens of thousands of reads. The data infrastructure, however, is thin. Basketball has tracking cameras, football has event-data providers feeding every possession, while Vietnamese table tennis still relies largely on a human recorder sitting beside the table. A rally lasts a few seconds; ball speed off a national-level male player's forehand loop is far beyond manual notation. Every domestic table tennis stat sheet has holes, and the professional question is not how to fill them but how to disclose them.

The annual season adds pressure. Readers follow every round and want to know who is peaking, who is fading, whose shoulder is overloaded by too much forehand loop volume. That demand is legitimate. It also creates a market where a writer without numbers is judged unqualified while a writer with numbers — wherever those numbers came from — gets shared.

Table tennis data splits into two distinct layers. The base layer is verifiable through rules and equipment: a game to 11 points, service changing every 2 points, a 40mm plastic ball, a table 2.74m long, a net 15.25cm high — all set out in International Table Tennis Federation regulations. Anyone can look it up, and because anyone can, this layer creates almost no differentiation for a piece. The observation layer is where value is born: points won on serve, successful receive rate by spin type, average rally length, placement distribution of the forehand loop. Nobody publishes that layer ready-made.

My method, developed over years of building models for the DataCourt podcast, is to code four fields per rally: server, spin type, placement, outcome. A rally enters the sample only when all four fields are complete; a rally missing a field is discarded — no interpolation, no guessing by feel. The consequence is a sample always smaller than the true rally count, sometimes down to 60-70%. In exchange, when the sample says a player wins 68% of points on short sidespin serves to the left corner, that number holds up under challenge. Double verification has two layers: one checks the provenance of the data, the other checks the internal consistency of the conclusion. Without the second, a writer is copying rather than analysing.

Based on my experience watching matches at domestic events, there is an under-reported trap: equipment. When a player changes rubber or blade, all prior spin and speed data loses comparability. I have seen form-comparison tables built straight across a rubber change with no note attached. The writer concludes the player is declining when he is simply adjusting to a different bounce. One line of equipment note can save an entire conclusion.

Here is the point worth dwelling on: data does not lie, but the story behind it is the truth. The same 55% receive-success rate, placed side by side, supports two opposite readings. Player A reaches 55% because he attacks first, accepting risk to seize control of the point. Player B reaches 55% because he pushes safe to the side, surrendering initiative and waiting for the opponent to miss. The stat sheet prints two identical numbers; the video prints two different fates.

I once built an injury-prediction model for a compressed schedule and sat on the draft for five weeks just to re-check the coefficients. When it ran, the prediction matched reality. The bigger lesson lay elsewhere: the delay nearly cost the piece its moment. Perfectionism is not delay, it is the final verification pass owed to the reader — but it only holds value when paired with a hard deadline.

Most of the profession believes an analysis piece without numbers is useless. I see it the other way: a piece that dares to write "insufficient data" exactly where data is missing is worth more than one that fills every cell with interpolation. The reason lies in what readers do with the information. A coach reads my work to decide whether to change a young player's serve patterns. Hand him a fabricated number and he changes the wrong thing while ignoring the right one. A wrong number in coaching costs far more than an honestly labelled gap.

One more counter-intuitive point: the silence of data is sometimes itself a tactical signal. A player with very few long rallies on file is not necessarily short of stamina. More often the opponent has read him and cuts rallies short with a short serve, forcing him to solve the point within two beats. When the long-rally sample collapses abruptly across three events, that is evidence the opposition has found a plan. Collapse is not instantaneous; it freezes quietly over three seasons — and in table tennis, three events are enough to drop a player out of the seed-protected bracket.

Revolutions always begin with a number nobody bothered to record. For Vietnamese table tennis, the most neglected number is the count of rallies that were never logged — the ones absent from every stat sheet, and therefore owned by no one. If each national event published one line noting the share of rallies excluded from the sample, fans would learn to read reliability before reading conclusions. Is a table tennis culture willing to publish the data it does not have a more mature one?