Missing Data: Vietnamese Sports Writing Its Own Unverified Miracles

Thiếu dữ liệu, Missing Data: Vietnamese Sports Writing Its Own Unverified Miracles In April 2026, in a small press room in Binh Duong, a veteran coach called my work "soulless statistics." He said it about ten minutes after his team lost 0-1 in a match where they held over 60 percent of possession and took nearly twenty shots. He added a line I have never forgotten: "Football is emotion, not Excel." I did not argue. I went home, opened the tape, and started counting. I counted for four months. One hundred and eighty-two V-League matches in the 2026 season, recording for each one how many passes the defending team allowed before intervening again, the metric European analysts call PPDA. The result made me sit still for a few minutes. Long An had the lowest PPDA in the league: 7.8. They barely pressed at all, inviting opponents to hold the ball freely. And yet across the whole season they conceded only 0.7 goals per match. That number told a story entirely opposite to what the terraces screamed every weekend. The stands called them cowards. My spreadsheet called them a team that knew exactly what it was buying with what it was giving up. Numbers never lie. We simply have not asked the right question. A career that began in a gap I was born in Korea, started out in 2026 as an esports athlete and then a tournament organiser, and drifted into media. In 2026, at twenty-five, I took a job at a new football outlet in Binh Duong. Eighteen years of watching this industry, and most of that time I have spent doing something very repetitive: reconstructing what happened on the pitch with numbers instead of with narration. Put differently, I am a Korean writing about Vietnamese esports for a Vietnamese audience. That position carries an advantage and a hazard. The advantage is that I see patterns insiders of a single market cannot see, because their own context blinds them. The hazard is that every sentence I write can become an insult if I forget that the audience here has its own history, its own memories, its own sleepless nights that no spreadsheet records. And in Vietnam, people in my trade keep colliding with the same thing: missing data. It helps to separate two kinds of missing. The first is missing because nobody measured. The second is missing because people measured but will not publish. The first is a problem of technology. The second is a problem of power. In the V-League both coexist, and they meet at one uncomfortable point: wherever there is no number, there is storytelling. And wherever there is storytelling, there is the word "miracle." Watch an ordinary V-League match on Vietnamese television. You will be told possession share, shot counts, foul counts, yellow cards. You will not be told which shots were real chances, who created the space for the shooter, where a team chose to lose the ball, and where losing the ball gets punished. You are served a meal with plenty of seasoning and no meat. The paradox is that raw numbers are plentiful while contextual numbers are close to zero. Possession share is a number with no standalone meaning. A team with 65 percent possession losing 0-2 did not play better than a team with 35 percent winning 2-0. Yet in most coverage, 65 percent is still presented as evidence of superiority and 35 percent is still read as evidence of luck. That is why I started with PPDA. The metric is simple: how many passes you allow the opponent before you win the ball back. Lower PPDA means you press higher and earlier. Higher PPDA means you sit deeper. At Long An in 2026, 7.8 was the lowest in the league, and the most naive reading is that the team refused to close down. The correct reading is that they chose to press somewhere else. Long An did not press in the opponent's half. They pressed in midfield, after the ball crossed the halfway line. They let opponents pass sideways and backwards, harmless passes, and when the ball entered the trap they had set, they cut it out and countered within three passes. The outcome was 1.4 goals scored and 0.7 conceded per match, a ratio that at the time only clubs with triple the budget could achieve. I wrote a piece headlined "Low pressing is not cowardice." It was mocked. A veteran coach told me statistics have no soul. Then a young assistant at Binh Duong called and asked whether I could build pressing maps for his club. That detail matters more than the whole article. The real argument was not between me and my critic. It happened inside a club, between two generations of coaching staff, and it began with a data cell nobody had thought was necessary. The V-League is a mess, but every mess has its own rules. Those rules are not in the top line of the table. They sit in metrics nobody sells tickets for: passes allowed before recovery, average defensive line height when possession is lost, duel win rate in your own third. Things that never make headlines but decide who gets relegated in October. Four times the spreadsheet made me shut up If you want to understand why I trust data to an uncomfortable degree, let me tell you about four times it embarrassed me. The first was the Long An piece itself. I arrived with the opposite hypothesis. I believed a deep-defending team was a negative team, and I wanted to prove it with numbers. After 182 matches, the numbers hit me in the face. The team with the lowest PPDA in the league was the least-conceding team in the low-budget bracket. My conclusion was arithmetically right and wrong in its question. I had asked "is this team negative." The right question was "what does this team trade away, and for what." That was the first and largest lesson of the trade: ask the wrong question and the prettiest data is useless. The second was Croatia, 2026. In 2026 I staked my entire career on a probability model named Croatia. After the World Cup quarter-finals in Russia, I was sent as an analysis reporter. I built a simple model with three variables: each side's average xG across the tournament, minutes played including extra time as a fatigue proxy, and penalty shootout quality if the match went that far. Croatia averaged 2.3 xG. England averaged 1.1. Croatia had played more extra time, and that variable was what made my colleagues laugh. In a stadium corridor, a colleague told me football is not mathematics, that Croatia were exhausted, that England's pace would crush them inside thirty minutes. I wrote the sentence in my notebook, because I believe a prediction only has value if it can be proven wrong. Croatia won 2-1 after extra time. The piece, "Goals from probability," was shared more than 10,000 times, and my editor gave me a column called "Seeing by numbers." But here I must be explicit, because people have since misquoted me. I did not prophesy. I did not see the future. I showed that a result called a surprise in fact carried a probability that was not small, and that what we call a surprise is largely a product of missing models. Croatia was not a miracle. Croatia was a well-managed variance. Let me translate that into ordinary language. Variance is the extent to which an outcome swings around its own average. A stable team has low variance: they win 1-0 or draw 1-1, rarely win 5-0 and rarely lose 0-4. A high-variance team spreads its results wide: some days they destroy, some days they are destroyed. Croatia in 2026 were a high-mean, high-variance team. High mean meant they generated clear chances at 2.3 xG per match, nearly double England's. High variance meant their matches tended toward extreme scripts: narrow wins, draws, extra time, shootouts. Managing variance means preparing for those extreme scripts. Croatia did not prepare to win 3-0. They prepared to play 120 minutes. They prepared for penalties. They prepared to go behind and come back. The team that prepares for the worst script survives the worst script, and in knockout football, survival is the entire game. That is what broadcasts call character. Character, in the language of data, is the ability to hold performance constant when environmental variables change. It is measurable. It is not a mystical quality falling from the sky. The third was the summer of 2026. The pandemic paralysed leagues, and the Bundesliga was the first major competition to return, in May, to empty stands. I had in my hands an opportunity no analyst can buy: a natural experiment at league scale. I took 252 Bundesliga matches from May to June 2026 and compared them against that league's own baseline in previous seasons. Home win rate fell from 43 percent to 29 percent. Away teams' distance covered rose by roughly 6 percent. Applause in an empty stadium records a truth nobody wanted to hear. Read those two numbers carefully. Home advantage, something every coach treats as self-evident as a law of physics, evaporated by 14 percentage points simply because human noise was absent. And away teams ran 6 percent more when nobody was jeering them. There are two explanations and I must say clearly that I cannot fully separate them. The first: referees are influenced by crowd noise, and when the crowd is silent, decisions become more balanced. The second: away teams lose psychological pressure, so they run more and play freer. Both may be true, in proportions my data cannot determine. This is why I keep repeating one principle: correlation is not causation. Empty stands accompanying a fall in home win rate is a correlation. The mechanism behind it is a different question, requiring a different and better research design, with control groups and controlled variables. I posted the comparison online. The Analyst shared it as scientific evidence for home advantage. That led to an invitation to collaborate with a European data platform. But what I kept from that summer was not the praise. It was the realisation that systemic shocks, whether pandemics, disasters or financial crises, are the only laboratories the universe opens for data people. And we usually miss them because we are too busy reporting on them. The fourth was Euro 2026, the tournament played a year late and still branded Euro 2026. I published a study of 342 penalties across five European leagues. In that sample, facing a right-footed taker, goalkeeper Gianluigi Donnarumma dived to his right 72 percent of the time. That is a very strong bias, and it is not a joke: it is a measurable, repeated habit large enough to create an edge. I predicted Italy would beat Spain on penalties. The piece was called fortune-telling. A week later Italy won 4-2 in the shootout and Donnarumma saved two attempts to his right. The article reached 1.2 million views. An international sports broadcaster hired me as a data expert for the 2026 World Cup. I tell these four stories to do one thing: show you that every time the spreadsheet silenced me, it did not make me wiser. It only showed me where I had asked a stupid question. Now back to Vietnam. What happens if I apply those four models to the V-League and to Vietnamese esports? The honest answer is that most of them cannot be applied, because the baseline data does not exist. For PPDA I need event data per pass, with coordinates and timestamps. The V-League now has part of this, but it is not fully published and not standardised across seasons. I can analyse one match. I cannot build a time series. And the time series is what tells you whether a team is improving or rotting. For xG I need a chance-quality model built on that league's own data. A model built on the Premier League and applied to the V-League will be wrong, because finishing quality, goalkeeper quality, pitch quality, even refereeing tendencies differ. A V-League xG model must be built from the V-League. Nobody has done that publicly and sustainably. For crowd data I need attendance, distribution and noise levels. We have tickets sold. We do not have noise data, we do not have block-level distribution, we have nothing with which to replicate the Bundesliga experiment at the scale of a Vietnamese stadium. For shootout data I need a database of every penalty in Vietnamese competitions over many years, with shot direction and goalkeeper dive direction. That dataset, if it exists, is scattered across dozens of personal hard drives and hundreds of unlabelled recordings. That is the third meaning of missing data: not nothing, but something nobody gathers, nobody labels and nobody is responsible for turning into a public asset. Now esports, where I work every day. Vietnamese esports has the inverse paradox. Here there is more data. Major titles publish post-game statistics: kills, deaths, gold, damage dealt, damage taken, fight participation. Publishers run APIs, statistics portals, near-real-time feeds. And yet the quality of public analysis in Vietnam remains low, because most published metrics are outcome metrics, not process metrics. Kill counts tell you a player finished a fight successfully. They do not tell you who created that fight, who forced the opponent to burn a long cooldown first, who stood in the wrong place in the first two seconds and forced the whole team to rotate. This is where I want to be blunt, because it is the biggest blind spot in both Vietnamese football and Vietnamese esports. Heat maps have become the new fortune-telling A heat map is the most visual thing data can offer. You look at a blue-and-red rectangle, you see a player spending time on the left flank, and you conclude immediately: this player favours the left. You see an esports player with dense points in the centre of the map and conclude: this player is a control player. Both conclusions may be true. Both may be meaningless. A heat map tells you where someone was. It does not tell you where they were asked to be. If a right-back repeatedly appears in central areas, there are two opposite possibilities. First: he is a player who drifts inside naturally, and that is his quality. Second: the team's structure broke down, he was forced to cover for a teammate, and his heat map is a record of collective failure. On the heat map, those two possibilities look identical. That is why I call the heat map the new fortune-telling. It gives viewers a feeling of understanding while in fact repeating positions without explaining roles. To turn it into a real tool, you must tag every action with a role: who initiated, who supported, who cleared the path, who created space. That tagging requires an expert viewer and cannot be fully automated. In Vietnam, the number of people who can do it fits on one hand, and most of them are not paid to do it. The same problem appears in injury data. In professional sport, medical confidentiality keeps fans and media in an organised fog. Clubs announce injuries when announcing benefits them. They stay silent when silence benefits them. In Europe the motive is sometimes measured in share price: a listed club weighs how much news of a serious injury will push its valuation down. In the V-League there is no share price, but there is something more sensitive: transfer value, contract negotiation and standing in the dressing room. An early injury announcement can depress a player's price in the next transfer window. A late announcement can wreck an opponent's plan. The result is that injury data in Vietnam does not exist as data. It exists as rumour. And rumour cannot be used to build models, because rumour has no reliable date and no reliable diagnosis. In esports the issue has a milder variant. Teams usually announce absences in general terms: health issue, personal matter, temporary break. Fans do not know whether that means burnout, a wrist injury, a mental health problem or a collapsed contract negotiation. Those four causes lead to four entirely different predictions about how long the player will be out and at what level they will return. Missing data here is not merely missing information. It is a mechanism that redistributes power: those who know hold the edge, those who do not can only speculate, and those who speculate are easy to lead. We think we understand the game until the spreadsheet opens our eyes. At this point I have to argue against myself, because otherwise I am just selling another faith. For years I dismissed the reactions of coaches and commentators. I thought they feared data because data threatened their position. That is partly true. But later I realised something more important: many things on the pitch are not in the data I have, and people who have worked in the game for decades know this through professional intuition. A coach sees his player lose focus after a family event. He sees a young defender afraid of long balls because he was beaten twice last match. He sees a midfielder with a sore ankle who still asks to play. My spreadsheet has none of those lines. My spreadsheet only has the results of those lines. Counter-intuition has value only when it explains what intuition has not yet explained, not when it denies intuition. If my model contradicts the observation of someone who has watched that team for thirty years, the high probability is that my model is missing a variable, not that he is blind. I write this not to appear humble. I write it because it is a professional rule. If you ignore the variables you cannot measure, you do not have a model. You have a summary of the things that are easy to measure. And that is exactly what is happening in Vietnam. We have a great deal of analysis labelled as data, but it is in fact the easiest data to grab: goals scored, kills, win rates, unbeaten runs. Everyone has these. They create no advantage for anyone. They create only the feeling of understanding. In Europe, the difference between a mid-tier club and a big club over ten years lies in whether the big club built its own data before it needed it. They do not start collecting when the season kicks off. They collect in silence, and when the transfer window opens they hold something nobody else holds. In Vietnam that gap remains wide open. A signal for the next cycle Over the last three seasons I have noticed a small but directional change: V-League clubs have begun hiring analysts part-time. The numbers are small, the quality patchy, and most of these people quit after one season because they are underpaid and given no authority. But the signal has appeared. In Vietnamese esports the same signal arrives from a different direction. Stronger teams have started appointing dedicated analysts, and a few have started archiving internal scrim data. This is where everything begins. A team that does not archive scrims has no baseline. A team without a baseline cannot tell whether it is improving or deceiving itself. I do not think this will produce a revolution in two years. Data revolutions do not happen like that. They happen slowly, through people who get fired, projects abandoned, spreadsheets left on the hard drives of those who have moved on. But I think I know what I will watch next season. I will watch average defensive line height when possession is lost. It is the cheapest metric to collect, needs the least technology, and says the most about a team's philosophy. A team that keeps a high line and loses the ball in midfield is a team that believes in winning the ball back quickly. A team that drops to its own penalty area and loses the ball there is a team that believes in defending as a block. Both can win. Only one fits the specific human beings in that dressing room, and that question is answered by data, not by inspiration. I will watch chance conversion rate in the first fifteen minutes of the second half in esports. That is the window where teams apply adjustments, and teams without data on that window tend to lose by repeating their first-half mistakes. I will watch how clubs announce injuries. If a club starts publishing expected return timelines alongside injury type, that is a sign it has understood that excessive secrecy is also a cost. And I will keep the Croatia story in mind. Not because it was right. Because it reminds me that a result called a surprise is only a result our models are too crude to predict. Eighteen years of watching sport taught me something deeply uncomfortable: "miracle" is the word we use for regions where data is still missing. When you start measuring, the miracle shrinks. Croatia shrinks into a well-managed variance. Long An shrinks into a deliberate low-pressing system. The crowdless Bundesliga shrinks into fourteen percentage points of home advantage created by human noise. Donnarumma shrinks into a 72 percent habit of diving right. The miracle does not disappear. It moves, migrating to the next region where we have not bothered to place a sensor. And in Vietnam, that region is still very large. The question I want to leave behind is not which team will win the title. It is this: of all the things you believe you understand about this sporting landscape, how many rest on observation, and how many rest on a table of numbers you read too quickly and never dared to check against its source.