The Data Blind Spot in Table Tennis: When Blank Spaces Get Filled With Stories

The WTT 2026 season entered its fourth week. I received a data file from an analytics partner. The column for "first-three-shot efficiency" across one round of play had been left blank — not a single number, not a single matrix, not a single player name. The sender attached a short note: "Data not in yet, handle later." I looked at that blank space for a few minutes, then closed the file. In the data analysis profession, a blank space does not mean "nothing to read." It means "something is waiting to be filled." Twelve years of tracking table tennis through numbers taught me one thing: the greatest danger is not in wrong numbers. The danger is in blank spaces filled with stories that sound too fluent. When data falls silent, people tend to write on its behalf — through intuition, through memory, through patterns seen elsewhere. The data systems of modern professional table tennis operate on an infrastructure far more complex than what television audiences see. Each WTT match is recorded across multiple layers: high-speed cameras capturing every rally, sensors on the table tracking bounce points, motion-recognition software dissecting each stroke into technical parameters. From those, indicators are constructed — win rate on serve, first-three-shot efficiency, average rally length, win rate in rallies longer than seven shots. But that infrastructure has fracture points. A tournament in a region with weak transmission. A camera system with synchronization failure. A round postponed for weather. A data collector switching software vendors mid-season. Any fracture in that chain is enough to create a blank space in the data table — and that blank space is usually not automatically marked as "unknown." It is simply empty. Under those conditions, the analyst's role is no longer to translate data into judgments. The role is to verify whether the data actually exists to be translated. This is the step many sports reports skip. They receive a data set, see a blank, and fill it with a judgment based on memory of the match — or worse, on expectations about the player. The result is an article that reads very convincingly, with charts, with terminology, but whose core is built on nothing. I once witnessed such a report. In 2026, an analytics site published an article about the pressing efficiency of a table tennis team at the Asian Championships. The article had charts, numbers, conclusions. The only thing missing was the data source. When questioned, the author admitted to having "estimated based on watching the match." A data analysis built on feeling — and no reader noticed until someone checked. The worry is not that a single article fabricated numbers. The worry is the mechanism that makes this easy and hard to detect. That mechanism has three layers. Layer one: unlabeled blanks. In most sports data systems, an empty cell is not distinguished from a cell with a value of zero. To a computer, both are null or zero. To a reader, both say nothing. But semantically, they are entirely different. An indicator equal to zero means "the player executed no such strokes." An empty cell means "we do not know what the player did." These two states lead to opposite conclusions, yet in the data table they look identical. In table tennis, this is especially serious for context-dependent metrics. Take short-serve win rate. If a player serves short 20 times in a match and wins 12, the figure is 60 percent. But if the system only recorded 8 serves, with the rest lost to camera error, that figure is meaningless. The problem is that nobody knows whether the remaining 12 exist. The 60 percent is published. The 60 percent is cited. The 60 percent becomes the basis for a judgment about form. Layer two: content pressure. The sports media industry runs at a pace that data cannot match. A match ends at 10 p.m. The analysis must be live by 11 p.m. In that hour, the writer must review the match, gather data, and build the piece. No time to wait for the data system to synchronize. No time to verify each blank. Under those conditions, the blank will be filled with what is available: memory of the match, feeling about form, and patterns seen in prior matches. This is where the danger shifts from "missing data" to "wrong data." A judgment based on memory is not subjectively wrong — the writer genuinely believes what he writes. But it is objectively wrong, because memory of a table tennis match is dominated by impressive rallies, not by statistical distribution. A winning backhand loop at the end of a set is remembered more clearly than ten scattered faulty serves throughout the match. When a writer fills a blank with memory, he inadvertently reproduces that bias in the analysis. Layer three: the self-reinforcing loop. When a wrong judgment is published with the appearance of data, it becomes a source for later articles. One site publishes an "estimated" figure. Another cites the first. A third cites the second. After three rounds, the original number — which was only a guess — becomes "common data." Nobody can trace it back, and nobody wants to, because the number now matches the expectations of the majority. In table tennis, this loop appears most clearly in "form" indicators. A player is described as "rising" after winning three straight matches. The judgment is repeated, cited, used as a premise for predictions. But three straight wins might be the result of three favorable draws, or three injured opponents, or simply random fluctuation in a small sample. No metric in those three matches is sufficient to conclude a long-term trend — but because the win rate displays clearly, it becomes the axis of the entire narrative. I call this phenomenon "the illusion of weighted probability." In table tennis, a player with a 70 percent win rate at one tournament may only reach 40 percent at another, because opponents differ, tables differ, balls differ. Without controlling variables, the 70 percent has no predictive value. In practice, however, that number is used as a fixed attribute of the player rather than as a conditional observation. The same happens with more detailed technical metrics. Take PPDA — passes allowed per defensive action — imported from football into table tennis analysis in recent years. The problem is that it was built for a sport with entirely different spatial and temporal structure. In football, PPDA measures a team's pressing intensity across the pitch. In table tennis, the concept of pressing does not exist in the same sense — there is no space to occupy, no pass to intercept. Applying this metric to table tennis requires a complete redefinition, and each analytics group defines it differently. The result is numbers sharing the same name but measuring different things. One group counts strokes before the opponent scores. Another counts strokes before the opponent executes an attack. A third counts time between points. All three call it PPDA. None is wrong — but none can compare its data to another's. And when numbers are cited without definitions, the reader cannot know what is being compared to what. This is another form of data ambiguity — not a blank, but a gray zone. Gray zones are harder to detect than blanks, because they look complete. A data table with every cell filled, every column present, every unit labeled — but the cells do not measure the same thing. The reader sees formal completeness and assumes substantive completeness. In modern table tennis analysis, gray zones are more prevalent than blanks. There is a fourth layer I only recognized after many years: the temptation of proprietary metrics. When everyone has access to basic indicators, competitive advantage shifts to proprietary ones — metrics only one group holds, built by undisclosed methods. These are often presented as "breakthrough findings" or "different angles." But because the method is undisclosed, no one can verify whether they measure what they claim. In table tennis, this appears in metrics like "psychological pressure index" or "emotional stability coefficient" — concepts that sound compelling but lack clear operational definitions. Such a metric might be calculated by counting how often a player displays negative emotion on court — based on video analysis by an unvalidated facial-recognition model. The resulting figure is then presented as an attribute of the player, not as a conditional observation from an unverified model. Players leave the court, spectators leave the stands, but data never leaves the game. For that reason, responsibility toward data extends far beyond a single match. A number published today will survive in next week's articles, next month's, next season's. It will become a premise for further analysis, be cited in debates, shape how fans perceive a player. If that number was built on a filled blank, the error does not stop at the original article — it spreads across the whole ecosystem. Tactics are what people draw on the blackboard. Data is what they draw on reality. But reality is not always ready to be drawn. There are matches where cameras fail, tournaments where systems desynchronize, metrics whose methods are undisclosed. In those cases, the analyst faces two options: acknowledge the limit, or fill the blank with a story. The second option is always easier. It requires no explanation, no waiting, no facing the emptiness. I once wrote an analysis of a tournament where most technical data could not be collected. I chose to publish what was available and mark clearly what was missing. A colleague responded that the piece "lacked a sense of completeness" — no charts, no decisive conclusion, no story. I agreed. But that incompleteness accurately reflected the state of the data set. A complete article about an incomplete data set would have been a wrong article. The real danger is not that the writer lacks information. The real danger is that the writer does not accept the lack of information. When faith in metrics becomes part of professional identity, admitting a data limit becomes harder than producing a substitute number. And in an industry where speed is prioritized over accuracy, the substitute number always wins. The irony is that in many cases, a clearly marked blank is more useful than a complete but ambiguous data table. When a report says "we have no data for this metric," it gives the reader honest information: there is a limit. When a report presents a number without definition and source, it gives the reader misleading information: there is no limit. Between those two, honesty about the limit is worth far more. My prediction model has no heart, and that is why it never gets hurt. But that is only true when the model is fed real data. A heartless model fed fabricated data produces cold, unshakable wrong conclusions. It does not waver, does not doubt, does not self-check. It merely reproduces the builder's bias in arithmetic form. So in every analysis project, I allocate a separate phase to verifying the existence of data before verifying its meaning. That phase produces no beautiful charts, no compelling conclusions, no stories. It answers one question only: was this number actually measured, or merely assumed? If the answer is assumed, the number never enters the analysis — however well it matches my expectations. I learned that principle from a small incident a few years ago. I was preparing a report on a young player's serving efficiency. The data table was complete, the metrics matched what I had observed watching the match. I had nearly finished the conclusion. Then I decided to recheck the origin of one metric — just one, not the whole table. It turned out that metric was inferred from an estimation model rather than direct counting, and that model had a wrong assumption about scoring. One wrong metric pulled three others into error, and the entire "serving trend" conclusion collapsed. The lesson was not to distrust models. The lesson was to distrust any number whose measurement origin cannot be traced. In table tennis, where hundreds of metrics are generated per tournament, tracing every number is time-impossible. So the more practical approach is to reserve rigorous verification for the metrics that will become the article's axis — the numbers used to draw conclusions. For auxiliary metrics, the standard can be lower, but the confidence level must be clearly marked. There are contracts mocked until the numbers tell their real story. The reverse is equally true: there are numbers celebrated until someone checks whether they actually exist. In recent years I have noticed a concerning trend in the table tennis analytics market: metrics multiply, definitions blur, and articles grow more confident. These three factors are not proportional to each other. More metrics does not mean better analysis. Sometimes that multiplication merely covers more blanks with more numbers. The empty stadium of 2026 taught me that table tennis is not just noise. When there is no crowd, no applause, no celebrations, one is forced to look at the structure of the match rather than its emotion. The same is true of data. When the noise of numbers is silenced, when beautiful charts and impressive figures are set aside, what remains is the most basic question: what do we actually know about this match? That question has no perfect answer. No data set is complete. No model lacks assumptions. No metric lacks limits. The only thing we control is our attitude toward those limits — accept them, mark them, and write around them rather than over them. The question I keep after every data table is no longer "what does this number say." The question is "does this number actually exist to say anything." When the answer is no, I choose to leave the blank standing. In a season where every match generates thousands of data points, the limits of data are not in what they measure. The limits lie in the user's decision: accept the blank, or fill it with a story that sounds too fluent. And in an industry racing for speed, that decision will shape not only the quality of one article, but the credibility of the entire analytics ecosystem for years to come.