Lessons From 4,812 Rows of Unusable Table Tennis Data

4,812 rows. That is the size of the tracking file I built over two weeks of logging table tennis matches, international and domestic. When I ran the first filter — dropping every row missing a match date, a tournament name, or both — the number of surviving rows was 0.

I had not mistyped a formula. I had not opened the wrong sheet. Most of the data I collected in those two weeks came from reports with no attribution, summary tables with no dates, and ranking-point quotations with no time stamp. A data row without a match date is not data. It is a sentence parked in a cell.

That failure was not planned. It taught me more than every spreadsheet I had ever built correctly.

A sport recorded from memory

Table tennis has an international ranking system run by the ITTF, with WTT launched in 2026 as the ITTF's commercial arm. Ranking points are calculated on a rolling 52-week window: the points from an event expire exactly one year after the match date. That mechanism turns the ranking into a record with an expiry date rather than a measure of form. The WTT calendar is tiered — Grand Smash, Champions, Star Contender, Contender, Feeder — and each tier pays points at a different rate, so the same win at two different events carries completely different weight.

In Vietnam, publicly available table tennis data is far thinner than in football. The national championship, the youth circuit, and the SEA Games are the three main sources. For players such as Nguyễn Anh Tú or Đinh Quang Linh, most international data is scattered across low-tier WTT events, where result sheets often publish only a final score without match duration or game count. Three sources with three different recording habits, merged into one table, produce a feeling of completeness without producing verifiability.

I built this tracking file to answer one narrow question: over the past 12 months, how many matches did a Vietnamese player actually play, against whom, and at which round did the run end. Three mandatory fields: date, tournament, opponent. I was not asking about form. I only wanted to know whether I had data at all.

The answer was no.

Three layers of data, three gaps

The points mechanism is the first gap. On a 52-week window, a player can hold a high ranking on last season's results while recent results sit far lower. My file recorded 1,638 of 4,812 rows with a complete match date, or 34%. Within those 1,638 rows, only 271 specified the round in which the player exited. Without the round, I cannot separate a qualifying-round loss from a semifinal loss. Both are logged as “lost,” and both are counted the same way.

The opponent field is the most frequently missing column, blank in 3,174 rows. In table tennis, the ranking gap between two players explains very little about the outcome. An attacking two-winged player facing a defensive pimpled-rubber opponent faces an entirely different problem than when facing a blocker who plays away from the table. A left-hander creates a different serve angle. A penholder has a different service rhythm. Lose the opponent's name and every downstream analysis loses its anchor: deciding-game win rate, point conversion when leading 9-7, third-ball win rate — all three need to know who is on the other side of the table.

The domestic pipeline is the third gap. The national championship and the youth circuit supply the sample, but results are usually published as a final ranking list without game-by-game scores. A ranking list tells me who won. It does not tell me how many games a semifinal lasted, which player came back from behind, or which group of athletes is genuinely improving season over season.

Based on my experience watching these matches, the biggest gap is always the round column. People remember clearly who won an event, remember more vaguely who reached the semifinals, and almost never remember who lost in the round of 32. Yet that bottom layer of data is what decides whether a player can hold a place at international level.

Those three layers do not stand apart. Together they produce a dataset with a full surface and an empty core.

Assumptions

Every conclusion here rests on my personal tracking file, not on official data from any federation. The file was built over two weeks, with 4,812 rows drawn from public reports. The subjective error sits in classification: some rows I flagged as “missing date” may have had a date had I read the original more carefully. If the true share of complete rows is above 34%, the direction of the conclusion about gaps holds, only its magnitude shrinks. And if an official, match-by-match data source for Vietnamese table tennis is published openly, this article will have to be rewritten from scratch.

The contrarian angle: correlation is not causation, but memory always claims to be

The 2026 World Cup taught me one thing: the model did not collapse — I was the one who believed it absolutely. I once ran a regression over 500 international matches and produced a 78% probability that a certain team would reach the semifinals. That team went out in the group stage. The error was not in the regression. The error was in letting a single indicator make the decision for me.

With table tennis, the same temptation appears in a subtler form. When match dates and opponent names are missing, writers tend to fill the vacuum with narrative: this player is “rising,” that player is “stalling,” this generation is “next in line,” that generation is “running out of air.” All of them are conclusions drawn from rows that do not exist. The habit reinforces itself: a claim repeated often enough becomes the reference point for the next claim, and a few months later nobody remembers where it started.

I read a player through thirty variables before I listen to a commentator. That principle only holds when those thirty variables actually exist. Data does not need my belief. Data needs my verification. When the verification comes back empty, the correct conclusion is “insufficient basis,” not “a trend is forming.”

There is a counterintuitive point here: an empty dataset does not mean the analyst failed. It is the first valid result, and it points precisely at what is missing — the thing a table full of wrong numbers would hide. It took me two weeks to collect 4,812 rows and two more days to accept that most of them were unusable. The two days were worth more than the two weeks.

What to watch in the next data cycle

Under the 52-week cycle, the points from each event expire exactly one year after the match date. That means a batch of points matures every month, and each time, the ranking tells a different story from its own version a month earlier. Readers need an anchor for comparison: an update date, a tournament name, and an opponent. Without those three, a ranking is just a string of characters.

Next time you read a line like “player X climbs the rankings,” try to find a date in it. If there is none, you are reading a sentence. And the question I keep for myself: how many times did I verify anything this week?