The Empty Spreadsheet and the Discipline of Silence in Sports Analysis

The Empty Spreadsheet and the Discipline of Silence in Sports Analysis Shenzhen, 3:12 AM The third monitor on the left runs a familiar stream: passes by zone, PPDA, duel win rate, the coordinates of every shot. The second screen holds the odds board of four Asian bookmakers, flickering by the second. The main laptop, the one I bought in the summer of 2026 and have never replaced, shows an empty spreadsheet. The connection is fine. The spreadsheet is empty because my filter just wiped the sample. The friendly-match set has a running density more than 25 percent below the seasonal average. The qualifying set sits in a different time zone and carries a different competitive motive. The domestic-league set differs sharply in physical intensity. Three sources, three contexts, none of them thick enough to build a baseline. On screen, every cell reads N/A. The colleague beside me, twenty-four years old and writing three times faster than I do, finishes a nine-hundred-word preview in twenty minutes. He has a conclusion. He always has a conclusion. My boss calls that productivity. My phone buzzes. A client writes: "Do you have a play for this one? I need it within the hour." I look at the empty spreadsheet, then at the clock, then type back a line I know will annoy him: "No side with enough support." He does not reply. I sit still. In this trade, silence is a decision, and every decision carries a price. This article is my accounting for that price. A Trade Pressured to Always Have an Opinion My name is Ngo Huy. I am twenty-nine, hold a bachelor's degree in sports journalism, live in Shenzhen, and work as a sports betting analyst. I started in 2026 as an esports competitor and tournament organiser, then drifted into media, then into data. For eleven years I have covered esports for the Chinese market and analysed European football for betting desks. Two markets, two languages, one habit: build the spreadsheet before opening my mouth. This trade has a peculiarity outsiders rarely notice. Readers do not pay for accuracy. They pay for certainty. A piece that says "I do not know" does not sell. A piece that says "this team wins 2-0" sells, even when it is wrong. People do not buy information; they buy the feeling that the world can be predicted. That incentive structure produces a new kind of analyst: someone who writes constantly, concludes instantly, and almost never looks back to check what percentage of predictions actually landed. I have worked beside these people. They are not liars. They simply have no mechanism in their heads for saying no. In Vietnam, this pressure takes its own shape. The Vietnamese sports audience reads fast, shares fast, and forgets fast. A three-thousand-word analysis may live twelve hours. A wrong prediction may live three years, preserved in somebody's screenshot. That structure does not reward care. It rewards volume. I am not writing this to complain about that structure. I am writing to show that it is quietly mispricing the entire analysis industry, and that the only way to last is to build a process capable of returning an empty result. The summer of 2026 taught me this unforgettably. Global football stopped. Tournaments were postponed. Stadiums emptied. My first three months at a betting company were three months with no match to analyse. Instead of waiting, I assembled data on 3,200 players from 2026 to 2026 and built a decline curve by age. The result landed outside my expectations: wingers lose an average of 12 percent of their running distance after age twenty-nine. When football returned, the company used that model to price the 2026 summer contracts. I took a position on a name everyone knew: Willian, a thirty-two-year-old winger joining a Premier League club on a free transfer. Pundits called it sensible, even shrewd. I called it a curve about to turn down. He could not meet the intensity. I won that bet. From that quiet summer I learned something that later became my entire method: when there is no new data, people read old data with a lenient eye. With no match to test against, every assumption looks correct. That gap is where errors breed. The Patch and the State of the Meta Start at the most foundational layer of any analysis: the state of the game. In esports, one update can invert the entire power order within two weeks. I once watched a patch change jungle respawn timers in a popular Chinese mobile MOBA, and every team's prediction model took three weeks to catch up. Those three weeks were three weeks of mispricing. People who understood made money in that window; people who did not called it luck. But there is a detail few outsiders notice. The patch notes are not the meta. The patch notes are the designer's intent. The meta is the actual behaviour of hundreds of thousands of players over the following two weeks. The two often diverge widely, and that divergence is where value sits. Football has no weekly patch, but it has equivalents, just on a slower rhythm. Five substitutions is a patch. VAR is a patch. Semi-automated offside is a patch. Each one makes a layer of old data partly meaningless and opens a new layer of opportunity for whoever reads the direction of travel. I hold a fairly hard line on five substitutions, and I have kept it across several seasons. It has not made football prettier. It has turned squad depth into a weapon and turned the last twenty minutes into a pure war of attrition. That means a team's most important indicator now sits on the bench, not in the starting eleven. Models that only read the starting lineup will miss that entire component. So what does an analyst do when a new patch has no data yet? The correct answer: do not build a new model, build a list of assumptions that need testing. That is why the first cell in my spreadsheet this week returns N/A. I do not know which way the new meta leans. I know it will lean somewhere, and I know I do not yet have enough sample to say where. There is a paradox here that took me years to accept: the strongest analyst during a transition is not the one who predicts correctly first, but the one who asks the right question first. Predicting early is gambling. Asking early is building infrastructure. Tournament Format The second layer is the frame the match happens inside. The same team, the same roster, can look like a giant in one format and an impostor in another. A multi-game series rewards depth and mid-series adjustment. A single game rewards chaos. A seven-match group stage rewards stability. A single-elimination bracket rewards one moment. In esports, the gap between a best-of-one and a best-of-five is large enough to change how you recruit. A team built for long series needs a shot-caller, a coach who can read opponents across multiple games, and a champion pool wide enough not to be locked out after two losses. A team that only plays single games needs none of that. It needs one plan and the confidence to execute it. Look at the 2026 World Cup to see how stark this is. Saudi Arabia beat Argentina 2-1 in a group-stage match that no model in the world predicted. I spent days reviewing roughly 2,100 Saudi running actions from three pre-tournament friendlies. They deliberately sat very deep in those matches. Against Argentina they pushed their line unusually high and caught Argentina offside ten times in the first half alone. What matters is that the result does not prove Saudi Arabia is stronger than Argentina. It proves that in a single-match format, a team perfectly prepared for exactly one opponent can produce a result a seven-match format would never permit. Anyone who uses that match as a foundation for judging Saudi's true strength has misread the format. Format also decides how I read schedule density. A team playing every three days while its opponent plays every seven is not in the same physiological state. Their spreadsheets look identical on paper and differ on the pitch. I have seen models accurate to the percentage point fail badly because they ignored the schedule variable. The schedule is not supplementary information. The schedule is part of the lineup. In Southeast Asian regional tournaments this problem is more acute. Schedules are often set late, changed late, and teams lack dedicated performance staff. Recovery capacity then becomes a variable that cannot be measured from public data. I am forced to flag those matches as low confidence from the outset rather than pretend I have enough information. Roster and Players The third layer is people. There are four things I always separate when judging a team: paper strength, role fit, chemistry, and bench depth. These four often pull against each other, and that is where the market errs. Paper strength is the easiest to measure and the easiest to misread. A team that collects four big names can carry an enormous aggregate transfer value and still lose to a side with no stars, simply because those four need the same patch of space. In esports, my model once rated an all-star roster 18 percent above its actual output, and the error sat in judging each individual correctly while judging the combination wrongly. Role fit is the hardest to measure and the most important. A good player in position A can become average in position B, not because he declined but because the space he needs no longer exists. In esports this shows most clearly in the shot-caller role. A player with exceptional reading ability can become useless if the roster will not listen, and a player of average mechanics can become a pillar if placed in the right structure. Bench depth is the most neglected variable. With five substitutions, the value of a substitute rises but nobody prices it. A player entering at minute 65 does not need to be as good as a starter. He needs to be good at one specific thing for twenty-five minutes. Traditional metrics cannot measure that kind of value, because traditional metrics were designed to measure a full match. On the night of the 2026 World Cup, I looked at the ball with different eyes. I was twenty, interning at a small tactical analysis site in Shenzhen, and I hand-calculated xG for France's twelve shots against Argentina in the round of sixteen. The result kept me up until morning: Kylian Mbappe generated 1.8 xG from just four runs behind the defensive line. I wrote a piece arguing he was rewriting the definition of a winger. My boss called it nonsense. A week later, a betting analyst shared it. What I learned was not that I was good. What I learned was that self-calculated numbers persuade more than borrowed numbers, because the writer is forced to understand every addition he performs. Since then, every piece of mine carries a table I built myself, even when it takes three hours to produce a table used in two sentences. Coaching staff deserves a mention, because it is the most neglected part of Vietnamese-language analysis. A team with a strong head coach but no performance specialist and no opponent analyst will collapse late in a tournament. This is a silent personnel risk: it does not appear in any metric until the team loses three straight matches at minute 80. Regional Map The fourth layer is geography. Esports splits into regions with sharply different styles, and those regions differ not only in playstyle but in how they organise, how they coach, and how they monetise. A team dominant in one region can fail in another not because it is weaker, but because it is optimised for a type of opponent that does not exist elsewhere. Football is the same, just at a larger scale. Europe's leading football nations differ in pressing intensity, in long-ball usage, in how much risk they accept near their own box. South America differs in tempo. Asia differs in physicality and tactical discipline. When an Asian team meets a European one, comparing spreadsheets directly almost always misleads, because the numbers were produced inside two different competitive ecosystems. There is one indicator I always check when judging regional strength: how many players or competitors are exported to stronger ecosystems, and what share of them survive there. Academy output only matters if there is a real outlet. A football nation that develops two hundred players a year but places only five in top leagues has a vanity metric, not a pipeline. Vietnam occupies an interesting position on this map, and I do not say that as a courtesy. Vietnamese football has a structural advantage many larger nations lack: low development cost and dense youth competition. Those are excellent conditions for producing players. But that advantage only converts into national-team strength if a data system good enough exists to filter who is genuinely improving. Without it, the output is a steady stream of players and no way to know who will break out. In regional esports, I have written extensively about talent movement between regions and the export of Chinese coaching models to Southeast Asia. One mistake recurs in such writing: copying the model wholesale without adjusting for culture, currency, and tournament infrastructure. A twelve-hour daily training regimen works in an ecosystem with academies, sports psychologists, and base salaries. It does not work where players still need side jobs. Same number, two outcomes, and anyone reading the number without the context will always choose wrong. Club Finance The fifth layer is money. In esports, money comes from four main sources: sponsorship, publisher revenue sharing, parent-organisation salary, and third-party tournaments. These four are not equally stable. Sponsorship depends on viewer growth. Revenue sharing depends on publisher decisions. Third-party tournaments depend on the betting market. When one of the four contracts, every team's model comes under pressure at once. I lived through the period the industry calls the esports winter, and the most striking thing was not the number of dissolved teams. It was the speed. An organisation can sign a player to a three-year contract, then fail to pay wages six months later, with no mechanism protecting that player. Football has transfer rules, a governing body, a sports court. In esports, most contract risk still sits on the player. With football, I read club finance differently. Transfer fees are a market-psychology indicator more than a quality indicator. When a player is bought for a record fee, the market is pricing commercial revenue expectations, not expected goals. I once valued a deal 40 percent below market because my model only counted on-pitch output, and I was wrong. The model was not technically wrong; it lacked a variable: the commercial value of the name. In a major-tournament season, that variable spikes, because a player performing well at a major can double commercial value within three weeks. That is one reason I never separate sporting analysis from commercial analysis. They are one system. A club buys a player not only to win matches but to sell shirts, sign sponsors, and open a new market. If I analyse only the on-pitch part, I am reading half a document. And for the same reason I hold a fairly uncomfortable view of how representation contracts currently work. When an athlete is locked into a long-term representation deal, their public voice tends to be flattened, because every statement must pass through a sponsor's communications department. Fans end up with a processed version, and personality, an athlete's single greatest asset, is replaced by a safe talking point. I do not need to name anyone to say this. Just look at how post-match interviews have grown more alike. Rules and Governance The sixth layer is the rulebook off the pitch. There are five clusters I always check before writing anything about a tournament. The first is competitive integrity: any sign of match-fixing. The second is transfer and registration rules, because an ineligible player can wreck a team's plan with no advance warning. The third is contract compliance. The fourth is minor protection, a major issue in esports with many remaining gaps. The fifth is governance disputes between publishers and organisers. Each cluster carries its own risk type. Integrity risk cannot be predicted, only detected. Registration risk can be checked in advance. Contract risk requires reading documents, not headlines. Minor protection risk usually goes unwritten because it has no pretty statistics. Governance risk usually surfaces only after it has already happened. In the Vietnamese market there is one more cluster I handle very carefully: the grey zones around betting. My trade is betting analysis, but analysis and instruction are different things. I can discuss probability. I do not tell anyone where to place a wager. That line is not a formality. It is the condition under which I get to keep writing. I once saw a very good analysis taken down only because its closing paragraph crossed that line. The writer lost three days of data and lost the channel as well. In this trade, knowing the rules is not an appendix. It is the opening section. And in a market where regulation moves faster than writers update, re-checking the rules before every piece is not excessive caution. It is basic risk prevention. Risk Profile The seventh layer is where everything converges. I split risk into six categories: competitive, financial, personnel, rules, public opinion, and systemic. These six are not independent. A personnel risk can become a financial risk within two weeks. A public-opinion risk can become a rules risk if it grows large enough to force a regulator's response. What I always do with a risk matrix is assign three values to each cell: level, probability, and mitigability. These three must stay separate. A risk with high severity but low probability must still be recorded, because it is the kind people ignore until it happens. A risk with high probability but easy mitigation does not deserve much writing time. My biggest lesson on risk came from the 2026 World Cup itself, and it had nothing to do with the scoreline. After Saudi Arabia beat Argentina, I asked the whole team to review all input data. We discovered that Saudi Arabia had deliberately played out of shape in pre-tournament friendlies to hide their setup. In other words, our input data was not randomly noisy. It was deliberately noisy. I told the team a line that later became a principle: old data is useless if the opponent is actively corrupting it. Immediately after, we rewrote the filter process, discarding any friendly match whose running density fell more than 25 percent below that team's own seasonal average. That is why my spreadsheet is empty tonight. The new process does not let me assume a small sample is representative. It forces me to accept that there are cases where I cannot say anything. And in betting, saying "no side worth taking" is a complete conclusion, not a surrender. Systemic risk is the kind I worry about most and write about least, because it has no concrete shape. A policy change in one country, a licensing decision by a publisher, a wave of investment withdrawing from a region: those three look unrelated on the surface, yet they can jointly break the business model of an entire tournament. An analyst who only reads matches will not see it coming. Public Narrative The eighth layer is the story the crowd is telling. Every major tournament produces a dominant narrative, and that narrative runs on a cycle. It is born from one good match, amplified by media, then feeds on its own spread. The problem is that the story and the truth do not move at the same speed. The story moves fast and lasts; the truth moves slowly and briefly. I usually measure the gap between market expectation and objective assessment at three points: team results, individual form, and transfer or injury-return moves. The largest gap marks where the market is most mispricing. In the summer of 2026, when Italy met Austria in the Euro round of sixteen and the crowd piled on Italy, I looked at two numbers: Austria's PPDA was only 7.8, meaning extremely aggressive pressing, and Italy's completion rate into the final third was just 21 percent. I recommended Austria plus one goal and under 2.5. The match ended 2-1 to Italy, but after extra time, and Austria held 48 percent of possession against a major side. I won the handicap bet. What I did not win was the argument. Many people still said Italy won. They were right about the result and wrong about the process. Readers of the scoreline could not see the stalemate I had calculated. That is the nature of public-opinion noise. It is not wrong. It is only slow. And here I have to correct an old habit of my own. For years I treated fan emotion as noise to be stripped out. That was methodologically wrong. Crowd emotion is a valid quantitative variable; it is simply a lagged one. If a very large share of fans believes a scenario, that belief moves ticket prices, viewership, and the pressure players carry, and eventually it moves results. Ignoring that variable is not objectivity. It is reading with missing data. Industry Transmission The final layer is the business itself. Everything in esports flows along one line: publishers upstream, clubs, tournaments, and streaming platforms midstream, sponsorship, derivatives, and mainstream cultural integration downstream. When a publisher changes a schedule or licensing terms, that wave takes six months to reach downstream. When an organisation goes bankrupt, the news reaches downstream in two days. I live in China and work for the Chinese market, so I hold a rare vantage point: I see this flow before it reaches Vietnam. That is both an advantage and a trap. An advantage because I can forecast. A trap because I easily forget that the lag is not a constant. Some changes take eighteen months to reach Vietnam. Some take three. Anyone who moves a model across without measuring the lag is applying an expired solution. At the same time, the data infrastructure of the two markets is fundamentally different. In China I can access granular data on every action, every draft pick, every minute of play, usually within minutes of a match. In Vietnam, most of that data must be rebuilt from video, by hand, with far greater error. Vietnamese analysts work harder to obtain a less accurate spreadsheet. That is a structural disadvantage very few articles mention, and it explains much of the quality gap between the two markets. That is why I always spend part of a piece clarifying which ecosystem I am talking about. An indicator that is correct in Shanghai can be meaningless in Hanoi. The Crowd Sleeps Inside Emotion; I Stay Awake With the Spreadsheet Now I have to say the hardest thing. The entire method I laid out above sounds very solid, and that is the problem. A process with nine layers, with tables, with a 25 percent filter threshold, with confidence annotations: it manufactures a false sense of safety. It makes readers believe the writer controls every variable. Nobody controls every variable. I built my personal brand on going against the crowd. That is an intellectually comfortable position and a psychologically dangerous one, because it creates a trap: once you are known for always pushing back, you start pushing back even with no reason. You defend an old argument not because data supports it, but because changing your mind would look like defeat. The biggest mistake is not placing a bet, but placing a bet with the crowd. Yet the second mistake, far less discussed, is betting against the crowd with no data fence behind it. Both are different ways of abandoning independent thought. I had to build a mistake log for myself. Every month I record the predictions I got wrong, specify which layer failed, and specify how I rationalised it while defending it. That log is not public, and I am not sure it ever will be. But it exists, because without a mechanism that catches my errors, I become exactly the kind of writer my work exists to oppose. Every match is a confession of probability. And the analyst must confess too, otherwise he is merely selling certainty. Where the Data Could Be Wrong I always close with a section on assumptions that could be wrong, and this time is no exception. The largest assumption in this entire piece is that thicker data yields more trustworthy conclusions than thin data. That holds in most cases, but fails in one specific case: when context changes faster than data collection. Under a new patch, a new tournament, a new roster, thick data may describe a world that has already vanished. A thin but timely observation can then be worth more than ten thousand lines of log. The second assumption is that an N/A state is honest. That is only true if my filter is unbiased. If my 25 percent threshold is set to discard the cases I do not want to analyse, then N/A is no longer honesty. It is selective silence. I have no way to test this objectively against myself. The third assumption, and the one I believe least: that readers actually want to know when I have nothing to say. Across eleven years in this trade, the evidence I have collected leans the other way. If any of those three assumptions is wrong, the entire argument of this piece needs rebuilding from scratch. The ball stops rolling, but the stream of numbers keeps flowing forward. The problem is that some nights the stream is slower than I want, and my job is not to push it faster. My job is to tell readers what speed it is moving at. Tonight that speed is zero. And the empty spreadsheet on my screen is not the failure of a process. It is the correct output of a process that is working. The next morning I reopened that spreadsheet and filled the first cell with a single line: "Sample insufficient. Wait two more rounds." That was the only prediction I dared publish that week.