The Empty Data Sheet in Vietnamese Table Tennis: A Gap That Does Not Fill Itself

In March, a nine-section report landed on my desk in Binh Duong. Every section had an assessment frame, a comparison table, a scoring rubric. Every data field read the same word: N/A. No player name. No event name. No ranking points. No head-to-head record. Not a single technical metric. The structure had been built, complete and orderly, to hold something, and the something never arrived. I have received thin reports before. Thin because the match was poorly documented, thin because the organiser published nothing, thin because I was lazy. This one was different. The extraction stage never ran, and everything downstream was generated as though it had. The report still looked polished. It was simply hollow. For someone in my trade, the problem is not one corrupted file. It is that the record-keeping layer of Vietnamese table tennis is thinner than in most other sports, and thinner quietly. A national championship can end up preserved as a photograph of a paper scoresheet. Qualifying results at a junior event sometimes exist only on the phone of the referee on duty. Point-by-point records, game by game, service series by service series, rarely become a machine-readable file. Meanwhile the WTT ranking system runs on a rolling 52-week window: old points expire automatically, new points must land in exactly the right place. A player without complete match records walks into a points-defence period with nobody, including the player, able to see where they actually stand. The three majors, the Olympic Games, the World Championships and the World Cup, generate pressure of a different kind from annual events. There, a single early-round loss can tilt a four-year cycle. To judge whether a player has genuinely improved, I need at least three layers: win rate against players from other associations, consistency at major events, and performance in deciding games. All three require raw data from individual games. When the first layer is missing, people substitute memory. Memory has a reliable bias: it keeps the best rally. This problem shows up across many sports. Table tennis suffers more, because each point is very short and the number of decisive points is small. The trap lives in the reflex to fill the blank, not in the blank itself. In 2026 I was 29 and published a home-built xG model for Becamex Binh Duong against Hanoi FC, predicting a 65% home win on the strength of superior possession. The result was 0-3. Hanoi held only 38% of the ball and took 11 shots from inside the box. I rewatched the tape for a month and found two missing variables: chance quality and the speed of central attacks. The more memorable finding was different. The blank in that model was never blank. I had filled it with an assumption nobody had tested. A year later I wrote a pre-final piece for the 2026 World Cup based on xG: France at 1.8 per match, Croatia at 2.4, a conclusion that France would struggle. The piece drew more than 200,000 reads. France won 4-2. My error was failing to adjust for knockout-round opponent strength, since Croatia had met weaker teams in the group stage. I sat down and wrote a 3,000-word self-rebuttal, published it on the same site, and attached the open dataset. Every model I own is built on mistakes that were once laughed at, the most honest foundation I have. The same thing happens in table tennis, at a smaller scale and with fewer witnesses. The decisive data layer in this sport is the first three shots: serve, receive, third ball. Without it I cannot separate a player who lost because serve quality dropped from a player who lost because the receive positioning was wrong. People will narrate the match by feel instead: this player looked stiff today, that one was in form. Those sentences may be true. They are simply not checkable. In a table tennis match, each game contains a few dozen points, and each point is a short chain of decisions. Skipping the record of that chain means skipping nearly the whole technical content of the match. What remains is the score, and the score does not explain why. In 2026, when competition stopped, I analysed 400 matches from the Bundesliga and K. League 1 to measure the effect of playing in empty stadiums. Home teams won 31% instead of 44%. I filed a 50-page report, met resistance, and held the conclusion because the sample was large and the measurement conditions were clear. The empty stadiums of 2026 proved one thing: data without context is only half the truth. In 2026, before the Euro final, I measured Italy's PPDA at 9.2 and England's at 13.5, and wrote that Italy would not let England breathe. Italy won on penalties. The piece persuaded not because I guessed right, but because I published the PPDA table for the whole tournament so readers could check it themselves. Those three cases differ in outcome and agree on one point: each had a foundation. A conclusion earns trust when match counts, measurement conditions and opponent names sit behind it. Remove those three and what remains is storytelling. That is why I run a gate on every report I accept: at least one named entity, at least three information points, and a source field with a real value. Below that threshold, the report goes back instead of moving forward into analysis. The rule sounds bureaucratic until you see a nine-section document of pure N/A that can still be read as professional judgment. My first reflex on seeing an all-N/A table is to declare that no data means no conclusion. It sounds like professional ethics. It also stops too early. An empty field is itself data. It tells you who has the resources to record and who does not, which events are systemically funded and which are not. In Vietnamese table tennis, a junior event without point-by-point records is not that event's private problem; it is a marker of how resources are distributed across the pipeline. Reading the blank as a finding is far more useful than filling it from memory. The opposite direction carries its own trap. I once believed transparency meant pushing the entire raw table into the article. That approach drives readers away without letting them verify anything, because nobody reads 4,000 unstructured rows. Transparency means keeping the decisive numbers and leaving the rest in the source file behind a link. And I hold myself to a rule: if new data contradicts a published piece, that piece gets corrected in public within 48 hours. Correcting is not humiliation. Defending a conclusion that data has already refuted is. Next cycle I am tracking one signal: the share of domestic events whose point-by-point records are stored as machine-readable files. If that share moves, every model downstream starts to mean something. If it stays flat, the most honest output remains a short paragraph stating that no assessable content exists, instead of nine beautifully structured sections. Data is not wrong, the reader is, and I have been that reader.