An Empty Data Sheet, and Conclusions That Still Get Published

On March 14, 2026, I received a spreadsheet from an analytics group I collaborate with on esports data projects. The sheet had eleven columns, enough to describe a professional match: possession duration, number of teamfights, gold difference at minute fifteen, teamfight win rate, major objective control rate, respawn counts, and four other advanced metrics. Every column was there. Not a single row of data. The team column read "TBD". The tournament column read "tier undetermined". The source column read "updating". The sample-size column read "not applicable". The sender attached one line: "Please write the analysis section for us, we'll add the numbers later." I read the spreadsheet three times. The first time to check for hidden sheets. The second to check the formulas. The third to make sure I wasn't misreading it. No hidden sheets. No formulas. Only the skeleton of an analysis, hollow inside. I replied with one sentence: "When there are numbers, I'll write." The answer came twenty minutes later: "But you write fast, the frame is already there." That is the sentence I have heard most in seven years of working in Busan. People do not lack data. They lack the time to wait for data. And in that waiting period, an article with a handsome frame can still be published, still reach the front page, still be shared a few thousand times — it simply contains nothing that can be verified afterwards. I was born in Vietnam, now live in Busan and report on esports for the Korean market. My daily work is reading metric sheets, cross-checking them against match recordings, and only then writing. The rhythm of this trade is compressed by three things running in parallel: the patch update cycle, the transfer window, and the major-tournament cycle. When all three overlap, newsroom pressure triples while verification time does not gain a single second. A meta update can flip an entire power ranking within two weeks. A change to an experience coefficient shifts the gold-timing benchmark of strong teams, dragging every prior conclusion out of alignment. If I take last season's numbers to describe the current season without stating the collection date and patch version, I have planted an error in the article that the reader cannot detect. Every meta update is a confession from the publisher — an admission that the previous balance state was broken — and each time, the old data block loses part of its value. Before discussing win or loss, I have to interrogate the numbers first. Where did they come from, how many matches are in the sample, what tool measured them, and what interest does the person behind them have in the numbers pointing that direction. The last three questions are usually skipped, and that is where the real damage sits. In 2026 I was nineteen, a second-year student in Busan. On World Cup night, I entered all twenty-three shots by the German national team against South Korea into an xG model I had written myself in Python. The output: 1.32 xG, zero goals, a 0-2 defeat. I cross-checked the footage and found that the naked eye is deceived through a very specific mechanism: eighteen of the twenty-three shots, or 78%, came from outside the penalty box. Watched with the eye, Germany looked like they were laying siege. Watched with data, they were standing outside the door throwing stones at a wall. On that Russian night, I saw a number that knew how to hurt. The defending champion went out not because of some invisible force. They went out because of a tactical choice that kept pushing the ball away from danger zones, and because of an opposing defence that accepted giving up that space. Two years later, in the 2026 season, K League 1 became the first football league in the world to resume in front of empty stands. My 2026 xG model began drifting systematically. I did not adjust the parameters. I collected 152 matches to find out what had disappeared. The home win rate fell from 46.2% in the 2026 season to 31.6%. The end result was a forty-page report concluding that every 10,000 spectators in the stands was worth roughly +0.08 expected goals for the home team. The 0.08 coefficient does not measure the silence; it measures what we lost. Nobody commissioned that report. I wrote it because if the foundation is wrong, every analysis built on top of it will be wrong too, and wrong in a way that is very hard to detect, because the charts still look fine. In December 2026, already a junior staffer at a newsroom, I was assigned to analyse Morocco — the first African team to reach a World Cup semi-final. I compiled the three knockout matches. Morocco conceded 71.6% possession, conceded only one goal, while opponents accumulated 4.02 total xG. The metric that made me stop longest was PPDA 25.1, nearly double the tournament average of 13.2. PPDA 25.1 — sitting deep is not a concession, it is stretching the pitch. Morocco deliberately let opponents pass in harmless zones, waited for the ball to enter a cuttable area, then counter-attacked. Korean media at the time called them passive. The metric sheet said the opposite: it was a calculated choice, executed consistently across three matches, with a margin of error small enough to defend the argument against rebuttal. I use these three examples in my internal training sessions for young contributors because they describe the same mistake at three different levels. Level one is an observation error: the eye sees a siege, the data sees long-range shooting. Level two is a model error: the tool is correct but the measurement environment has changed. Level three is an interpretation error: the numbers are right but assigned to the wrong cause, turning into emotional judgement dressed as science. In esports all three errors appear more often, because the patch cycle is far shorter than a football season. A team wins the spring split with a major-objective control style at minute twenty. Three weeks later, the publisher adjusts the stats of a few champions and one neutral objective. That style loses value. But if I write an analytical piece based on spring-split data without clearly stating the patch marker, I am selling readers a map of a city whose streets have been rerouted. The same thing happens with roster analysis. People like to say a team "lacks bench depth" after one loss. That phrase measures nothing. What is measurable is the minutes played by the substitute line-up over the last ten matches, the substitution rate between minutes sixty and seventy-five, and the metric gap between starters and substitutes by position. Without those three numbers, a claim about roster depth is just another way of saying "a feeling". In 2026 I turned twenty-five. The Morocco piece connected me with a sports data company in Lisbon. From that source, I discovered a Korean midfielder at a mid-table club had played only 564 minutes the previous season, far below the 1,200 minutes recorded in his contract. I sent his agent a six-page metric report. On June 8, 2026, I was the first to report the loan deal with a €2.8 million purchase option. A transfer fee does not measure talent; it measures the buyer's hunger. The €2.8 million figure only means something beside the 564 actual minutes and the 1,200 contractual minutes. The agent later told me they trusted me because I brought numerical evidence, not emotional judgement. That is the entire secret of this trade: make the other side see that they do not need to believe you, only to be able to check you. The logic frame I have used for every transfer story since then has four steps: hypothesis, data, source, probability. I removed vague phrases like "declining form" and replaced them with "minutes played down 41% versus last season". This way of writing is slower, drier, and sometimes pushes a piece down the trending list. In exchange it buys something I need more than pageviews: verifiability. Based on my experience following matches across many consecutive seasons, I believe the biggest problem in esports analysis today is not a shortage of data. Public statistics platforms are already good enough for anyone to build a very professional-looking metric sheet. The problem is that this industry rewards the shape of analysis rather than its veracity. An article with a headline, numbers, charts and a decisive conclusion gets shared more than an article saying "the sample is still small, not enough to conclude". That is the most serious blind spot, and it is counter-intuitive in exactly one place: an honest empty sheet is worth more than a full sheet with unclear provenance. An empty sheet forces the writer to wait. A full sheet with blurred sourcing lets the writer issue a conclusion immediately, and worse, lets the reader believe that conclusion has support. The error in the second case is not in the number. It is in the belief. I have also witnessed another variant: data analysts are increasingly moving close to the locker room. They hold club contracts, have access to training data, sit in tactical meetings. That access is a genuine asset. But their conclusions often detach from the team's actual rhythm, because the model does not know that a player has just been through three weeks of sleeplessness, that a position changed role last week, that a teamfight at minute ten was cancelled for a reason nobody wrote in the minutes. The data is not wrong. The data is simply missing the slice of life it was never designed to record. So when I write about a defensive team, I do not use the phrase "pinned back". I replace it with "deliberately sitting deep". This change is not linguistic courtesy. It forces me to prove the intention behind the behaviour: if that team sits deep in an organised way, the defensive line height should fall steadily along script, intercepted passes in midfield should rise, and opponent xG should cluster in low-danger zones. Without those three markers, they really were pushed back, and I must write exactly that. Sitting deep in this craft is a similar choice. When a hot event breaks — a controversial transfer, a national team eliminated in the group stage, a patch that flips the meta right before a major — I choose not to write in the first twenty-four hours. I use that window to collect, cross-check, and record the limits of the data. Readers see me as slow. I accept that, because most serious mistakes in this trade are produced in the first twenty-four hours, before a source has been confirmed a second time. During a major-tournament cycle the pressure is heavier still, because national-team emotion compresses into a single block. Fans want to know immediately who is stronger. Newsrooms want the piece before rivals. Platforms want content before the ball rolls. In that state, a missed penalty at minute eighty-eight is always explained by nerve, by mentality, by words that cannot be measured. I have no right to deny those factors. I only have the right to say that without behavioural data or pressure metrics to prove them, I will not put them in the piece as a cause. What I want to leave behind after all of this is not a rule. I want to leave behind a habit. Before sharing any metric sheet, ask three questions: how many matches is this from, which patch was it collected on, and does the measuring tool lean toward one team? Those three questions cost less than an argument, and they change the conclusion in more cases than people expect. The spreadsheet of March 14, 2026 is still in a folder of its own. I have not deleted it. It is evidence that this work does not begin at the analysis section, but at the source column — the column someone left blank, believing nobody would look. Whoever writes next for that group may already have filled in the analysis section. I genuinely want to know whether, reading it again three months later, they still believe what they wrote.