An Empty Spreadsheet and the Voice of Honesty: A Tennis Analyst Learns to Say 'Insufficient Data'

In the summer of 2026, in a small office in Liverpool, the screen in front of me showed 71.4%. Spain's possession share against Russia in the World Cup round of 16. I typed that figure into my prediction column, added a short line reading "Spain win," and sent the report off. Four hours later they left the tournament on penalties, losing 3-4. Across those 120 minutes, Spain completed 1,029 passes and generated a mere 0.9 xG. For a full week afterwards I rewatched every phase of play, and what struck me was not the model's margin of error, but the silence of a data column I had left blank without realising it. I bring up that old story because last week I was handed an empty input dataset: a single label, "tennis," and not one information point attached to it. No player name, no tournament, no timestamp, no source. For someone who does this for a living, it is the most awkward moment there is, and also the moment that lays bare the nature of the work most clearly. What do you do when the spreadsheet is empty? The answer seems easy, yet it is the line between an analyst and a fabricator. My job is to turn thousands of raw figures into a story that can be checked. Every day the system runs across hundreds of matches, stripping out every scoreline, every serve, every net approach, then sorting them into meaningful groups: player group, tournament group, surface group, phase-of-season group. The process resembles a water filtration line. Raw water pours in, clean water flows out. But one thing outsiders rarely notice: clean water and an empty glass look identical if you only look at the glass and never at the pipe. That is the trap. An empty input dataset can still look formally "valid." It has a domain label, it has field structure, it has the correct format. It is missing exactly one thing: content. And if I am the reader of the result and never question it, I will assume "nothing worth reporting," when the truth is "nothing was reported to me." Those two statements are worlds apart. In sports analytics, and especially in tennis, the gap between "no data" and "data equal to zero" is one of the fatal errors. A player with a 0% second-serve points won rate in a match almost never happens, but if he served only one second serve and lost that point, the figure is still 0%. Look at the table and the two situations are completely different; look at the number and they are identical. Without a denominator, a numerator says nothing. So the first rule I set for myself is this: when a dimension of analysis lacks input, I must write exactly the words "insufficient information," and I am not permitted to fill the gap with a default value. It sounds simple. In practice, the pressure to "reach a conclusion" is strong enough that people usually choose inference. And inference, in this profession, is often the first step toward fabrication. Imagine I had to describe the playing style of a player whose name I do not have, whose match I do not have, whose surface I do not have. I could write a flourish about "nerve at the decisive points" or "adaptability to a fast court." No one can verify it, because no one knows who I am talking about. But precisely because no one can verify it, it is the perfect lie. I call it a structured lie: right format, right terminology, wrong fact. I nearly fell into that trap myself. In 2026, when the pandemic emptied the stadiums, I worked as a data analyst for a tactical consultancy. The Merseyside derby in June 2026 finished Liverpool 0-0 Everton. I compared Liverpool's PPDA before and after crowds returned, and watched it jump from 9.8 to 11.5. Their high-intensity running dropped 4.3% in a noise-free environment. For one week I considered writing a neat conclusion: "The crowd decides Liverpool's pressing intensity." Then I stopped. I had one match, one context, a sample far too small. I did not write it. That was when I learned that honesty in analysis lies not in finding a pattern, but in knowing when you are not yet entitled to declare one. Old data is not wrong; it is only that I once placed it on the operating table in the wrong season. The same metric, set side by side in a season with crowds and a season without, tells two different stories. If I forget that context, I will tell both of them wrongly. Since then, every analysis I produce carries a "context declaration": which surface the match was played on, at what stage of the season, with or without a crowd, and what state the opponent was in. Without that declaration, no matter how beautiful the number, it is a bare number. And I stopped presenting bare numbers long ago. In 2026, I was assigned to analyse Leicester City's run of 15 poor matches after their FA Cup triumph. That squad had seven injured centre-backs, Jonny Evans missing 12 matches among them, and their expected goals conceded rose 24%. At first the public said simply: bad luck. I did not accept that explanation. I dug into the centre-backs' running distances: 8.2 km per match on average, but falling 12% after every match played less than 72 hours apart. The result was a metric I proposed called "expected injury load," and the company adopted it. An injury chain is not a curse; it is a map revealing the depth of a system being eroded. If you want a purely tennis example, look at Roland Garros, where Rafael Nadal won 14 titles across his career. No statistics table fully explains that figure once it is detached from the clay surface, from the five-set format, from the two-week rhythm of play. Every record is a chain of denominators, not a single number. With the current generation, from Daniil Medvedev, Jannik Sinner, Carlos Alcaraz to Novak Djokovic, the same serve metric placed on a hard court and on clay tells two different stories. A reader only understands it correctly when they know which surface, which round, and which phase of the season the number belongs to. What I want to say to Vietnamese tennis readers is this: most of the analysis you read daily fails not for lack of numbers, but because the writer fills the gaps with assumptions and never marks them as assumptions. A player losing a match because of "weak mentality" is the most convenient and most worthless conclusion, because it leaves no trace to verify. By contrast, a sentence like "he lost the third set 0-6 after first-serve percentage fell to 41%, down 23 percentage points from the first two sets" is a piece the reader can check by rewatching the match. I do not say this to criticise anyone. I say it because I made that mistake myself, and the cost of a wrong conclusion in the sports-data environment is not small. Readers trust the analysis, place bets, or simply carry a wrong judgment into their conversations about the player they love. A wrong judgment that circulates long enough will dress itself in the appearance of truth. And by then, no one remembers it came from a data gap. This is where I want to return to that empty dataset. It is not a rare phenomenon. In fact, it is the most common and most dangerous error class in a sports-data analysis chain: an empty result, correctly formatted, passed downstream without an error flag. The recipient has no way to distinguish "the article contained no noteworthy event" from "the system dropped the entire content." In both cases the output is identical. But the handling is entirely different. In the first case, you conclude as normal. In the second, you must stop and trace back. Error is the least likeable friend, but the only one who never lies to me in the meeting room. An empty dataset does not shout "I am empty." It simply sits there, silent, waiting for you to fill it with an inference. And if you fill it, you have volunteered to turn a gap into an illusion of knowledge. There was one small detail in that input I latched onto: the domain label "tennis" survived intact. That means the classifier recognised the topic. It did not fail at the root. The failure occurred further down, at the information-extraction stage. If the classifier ran correctly while the extractor returned nothing, it is quite possible the original text still exists, only that it was not converted into data points. That is a diagnostic tool, not an excuse. I picture three scenarios. First, the raw-text fetch failed, returning an empty or paywalled document. Second, the parser produced output but could not serialise the information-point array. Third, the field names between the two system tiers do not match. The third is somewhat the most plausible, since the domain label survived while the entire content field vanished, a sign of a schema-mapping error rather than a total data-fetch failure. But I must be clear: this is a hypothesis. I cannot verify the internal mechanics from an empty output alone. And this is the biggest lesson, the one I want tennis readers to remember: the silence of data is not a finding. It is a signal needing investigation. In tennis, this is equivalent to looking at a player's stats table and seeing a blank second-serve points won rate, then concluding immediately that he serves brilliantly or disastrously. Both conclusions are wrong, because you are judging a number that never existed. Someone will say: but if you do not reach a conclusion, you are not doing the analyst's job. I think the opposite. The greatest value of an analyst is that he dares to say "insufficient information" while the whole world around him demands a tidy answer. Refusing to conclude may look like weakness, but it is the highest form of protection for professional credibility. An analyst who is wrong once loses the right to be believed the second time. At major tournaments, this pressure is even greater. Fans follow their national team, wanting their belief confirmed, wanting to hear their side is stronger, that the defeat was only an accident. A data analyst standing between those two emotional currents has two options: say what people want to hear, or say what the data permits. For years I have chosen the second, and I know it does not bring rapid fame. But it lets me sleep after every report. At industry level, this is not merely the story of some analytics office. It is the story of how we consume sports information as a whole. When data is marketed as something objective, readers easily forget that data too is selected, framed, filtered through many layers of machines and people. A blank cell in a table may be the result of a technical accident, or of a decision not to put that information in print. Both can happen, and readers have the right to know which case they are facing. I return to the opening question: what do you do when the spreadsheet is empty? My answer, after years of stumbling, is three steps. One, delineate the emptiness: which stage it sits in. Two, distinguish a broken system from an empty source. Three, and most importantly, write in your own words that here there is insufficient data to conclude. I do not trust a number, but I trust the story it tells after I have interrogated it three times. And if after three interrogations it stays silent, the most honest thing is to admit: I have nothing to say. There is a paradox in this profession that I think any tennis analyst encounters. The more you understand data, the more you see its uncertainty. The more you see its uncertainty, the more cautious you are when declaring. And the more cautious you are, the more reserved you feel to the reader. That is the price of honesty in an industry that runs on manufactured certainty. But I accept that price, because what I am protecting is not personal reputation, but the reader's trust in their own capacity to understand this sport. And here is the most counter-intuitive thing I want to say, the one many in the industry will not enjoy hearing: a good data system is not the one that produces the most conclusions, but the one that speaks up to refuse a conclusion when the data is insufficient. We usually measure a model's value by the number of predictions it generates. I would measure it by the quality of the gaps it dares to leave empty. A model that invents a prediction for every case sounds powerful, but in truth it hides its ignorance behind confidence. A model that says "I don't know" exactly where it does not know is far more trustworthy. That is why that empty dataset, useless as it looks, is a value test. It shows me the distance between a disciplined analyst and a word-pumping machine. The word-pumping machine will look at the gap and fill it, because it exists to produce output, not to seek truth. The disciplined analyst will look at the gap and stop. That difference is not in expertise. It is in professional dignity. In tennis, where every point is the result of dozens of variables acting together within seconds, the temptation to fill gaps is far stronger. People want an explanation for every defeat, a reason for every missed serve, a label for every player. But the nature of this sport contains randomness. A ball clipping the line, a sudden gust, a shout from the stands, any of these can swing a point, a set, a match. If an analyst erases that randomness in exchange for a tidy conclusion, that person is no longer analysing tennis. That person is writing a novel dressed in numbers. So which signal should be tracked in the next round? Not which player wins the next match, but how the parties handle data gaps. A transparent system will report an error when extraction fails. An honest system will write "insufficient information" instead of filling a default value. And if you read an analysis where every gap is filled with suspiciously smooth ease, be suspicious. Perfect smoothness in sports analysis is usually the mark of a neatly arranged lie, not of a thoroughly understood truth.