The Silent Gap in Professional Table Tennis Data

In my personal data warehouse in Da Nang there is a folder named "blank". It holds table tennis matches I watched from start to finish but could not record a single metric for. The input source — the organizers' score sheet, a WTT statistics file, or simply my own handwritten log — returned an empty result. This kind of failure is more dangerous than an obvious error. A spreadsheet with a broken format is immediately visible and fixable. A file with the right header, the right sport label "table tennis", but an empty body is easily misread as "nothing worth discussing". Both lead to the same outcome: the analyst sits in front of the screen and starts inventing what the data never said. An amateur spreadsheet taught me that data does not need to be flashy, only correct. But "correct" is not merely a number matching reality. It is also knowing clearly when a number is absent — and refusing to draw a conclusion when it is absent. Of the ten V-League matches I logged by hand at seventeen, what cost me the most time was never the beautiful passages of play. It was the cross-checking stage. A counter-attack by SHB Da Nang that I marked "successful" on the left side of the pitch ended, on review of the footage, with a misplaced pass. The figure of 15% misplaced passes in the final third became the foundation for everything I wrote afterwards — because I had to check it three times before I dared put it on the board. The Da Nang data warehouse taught me: patience is the easiest algorithm to write and the hardest to run. Table tennis is even harsher than football here. A forty-minute table tennis match can contain three hundred rallies, each under three seconds. By mid-rally, the human eye can no longer separate topspin from sidespin, so if the input data breaks, no footage will save you. Recent WTT Champions events make this picture clearer than ever. WTT's Stats+ system publishes placement, win rate after serve, and efficiency in rallies of four shots or more. But the quality of those numbers depends on a chain of links: high-angle cameras, placement-recognition software, and an operator re-keying the missing scores. One silent link and the whole data chain becomes a zero. At the quarter-finals of a recent WTT event, I spent three days trying to reconstruct the "points won per total service points" metric for each top player. The initial result looked beautiful. One player in the world's top group reached 68% — about twelve percentage points above the tournament average. Looking only at that number, the conclusion would be: his serve is at its peak. But when I split the data by game, the picture collapsed. Games one and two were fully recorded; games three and four were blank. That 68% was calculated on half a match, and the other half — the two decisive games — went completely unrecorded. The beautiful number became a deceptive number, only because I nearly read it as a complete reality. This is where table tennis analysis must learn from football. People speak of xG, of transition metrics, as if they were truths. But anyone who has run a data warehouse knows: what determines the quality of an analysis is not the prettiest metric, but how many missing rows you have. Since then, every table I build has two extra columns no magazine prints. The first is the number of rallies recorded. The second is the estimated number of missing rallies, based on footage duration multiplied by average tempo. If the missing rate exceeds 15%, I mark it red and write no conclusion based on that table. That 15% rule did not come from a manual. It came from the Da Nang warehouse itself, when I found that a misplaced-pass rate above 15% accompanied two defeats in ten matches — a correlation strong enough to make me believe I was seeing a tactical problem rather than a random sample. To grasp the scale, take a concrete example. A player serving a short sidespin into the middle of the table wins the point directly about 30% of the time at the elite level. If the system misses two of ten such serves in the decisive game, the displayed rate can jump from 30% to 37% or fall to 23% — depending on whether the missed rally was a win or a loss. Two lost data rows, and we already have two opposite stories about the same player. Every player is a notebook; only the one willing to read sees the final line. In table tennis, the final line is usually the blank one. Top players like Fan Zhendong or Wang Chuqin are analysed down to the millimetre of spin. But when their service data is missing in the decisive game, all we are left with is feeling — and feeling cannot be reproduced for verification. I once tried to test this systematically. Of about fifteen WTT matches where I collected complete raw data, four missed more than 20% of rallies in the decisive game. In all four, the win rate after serve computed on raw data diverged from the rate recomputed with footage by at least seven percentage points. Seven percentage points in an elite table tennis match is enough to break any prediction. The analyst's first reflex on seeing empty data is to compensate with observation. I once believed that. "If the machine cannot record it, my eye can." But after logging a few games by hand and checking them against footage, I realised the human eye's error at fast rally pace is far too large. A rally I labelled "won after serve" turned out to have been decided on the fourth shot — outside the definition I was applying. That means when data breaks, filling it with observation does not create new data. It creates a different kind of data, a blend of the measured and the guessed, with no label to distinguish them. The most dangerous silent failure is not an empty file. It is a file that looks complete while half of it is guesswork typed at midnight. I do not believe in fate; I believe in correlation coefficients. And a correlation coefficient is only trustworthy when the sample is recorded, not when it is recalled. This is the line many table tennis analyses on social media cross without noticing: they cite numbers rewritten from memory, then use them to assert a tactical trend. This risk does not sit with the players. None of them loses a point because of an empty data column. It sits with the readers, who believe a printed number has been verified. In table tennis, where a single point can swing a whole match, a seven-percentage-point error is enough to turn a winner into a loser on paper. What I take away is not to abandon data and return to feeling. On the contrary, I believe in data discipline more than ever — but a discipline honest about its own limits. A good data warehouse is not the one with the most metrics, but the one that says clearly where it is silent. For professional table tennis, the signal worth watching in the coming months is not which player is finding form. It is whether WTT events will publish their own data-missing rates. When a system dares to state "we are missing 12% of rallies in this game", it does not weaken itself. It is teaching readers how to read. As for me, the "blank" folder will remain. I keep it as a reminder that the most honest answer to an empty file is sometimes simply: not enough to conclude. Fans remember player names, I remember contract expiry dates — and also the days my data expires, because who knows, next time, when it is fully recorded, the story may be different.