The Empty Spreadsheet: What Vietnam's Numberless Esports Analysis Actually Costs

2:47 a.m., February 12, 2026. I opened a report file that had just been pushed back to me from the system — the kind of file I receive once a week — and saw eleven consecutive blank rows. No tournament name, no team name, no player name, no patch number, no publication date. Every substantive field carried the same label: N/A. The operator at the other end of the pipeline had done one thing I did not expect. He had not filled in the blanks.

Three numbers to begin, as usual.

First, 80%. That is the salary-to-revenue ratio many mid-tier esports clubs touched during the 2026–2026 period, according to estimates widely circulated in the industry and confirmed by the finance directors of several large organisations in public interviews. Not a pretty number, but it exists because somebody sat down and counted.

Second, 43% down to 31%. The home win rate in the Bundesliga before and after football resumed behind closed doors, drawn from 250 matches I collected myself in 2026. Average goals per match fell by 0.4. With no crowd, home advantage largely evaporates.

Third, 0. The number of information points extracted from the analysis I was holding.

The third number is the one that made me sit back down. The first two are data. The third is an event.

The spreadsheet is an altar, and I offer myself to every number on it. But an empty altar gives nothing to offer. The story I want to tell tonight is not about a broken system. It is about what happens next, on the far side of the pipeline — where an empty analysis, instead of being thrown away, gets filled with prose that sounds entirely reasonable. And in Vietnam, where I have watched esports from both sides of a border, that habit has become a business model.

Data context: which tournament, which server, which crowd

Before any conclusion, I have to rebuild the frame. This is a rule I set for myself in 2026 and have never broken: no number without its environmental conditions. For esports, that frame has four layers.

The first layer is game identity. League of Legends, Dota 2, CS2, Valorant, Honor of Kings, Arena of Valor — each title has a different patch cadence, a different tournament system, and business logic that does not transfer. Riot Games patches every two weeks. Valve patches rarely and usually around Majors. Tencent operates on a seasonal structure. A conclusion drawn from Arena of Valor's patch cycle cannot be applied to the VCS, even though both are MOBAs with Vietnamese audiences.

The second layer is server and version. Professional play runs on the tournament server, not the public server. The gap between those two versions determines which meta a team practises on and which meta it competes on. There have been tournaments in which the champion was simply the team that adapted fastest to a change that appeared on the tournament server three days before groups.

The third layer is the crowd. The Bundesliga 2026 lesson does not apply only to football. When the VCS moved to empty-arena play during distancing periods, pace-of-game indicators shifted in less predictable directions, and anyone comparing data across the two periods without flagging the crowd condition is comparing two different things.

The fourth layer is schedule density. A team playing four matches in six days is not the same object as a team playing four matches in four weeks. Distance covered, pressure applied, overtime win rate — all of it is distorted by the calendar.

These four layers are necessary conditions. Miss one, and every conclusion must be capped at low confidence. Miss all four, and what you are writing is not analysis. It is prose.

And here is the point I want to state up front, because it is the spine of this entire piece: when data is absent, the default behaviour of the content market is not silence — it is filling. Silence does not sell advertising. A line reading N/A generates no shares. But a paragraph about "a shifting meta", "roster depth" and "a dip in form" does.

Layer one: what gets extracted

That empty report was not a rare accident. It was a diagnosis.

Picture the information pipeline as a three-storey building. The ground floor is extraction: pull raw text from a source and break it into information points — names, numbers, dates, statements, events. The second floor is identification: turn those information points into named entities — which team, which player, which tournament. The third floor is interpretation: attach tactical and economic meaning to those entities.

When the ground floor returns nothing, the two floors above have nothing to stand on. And the notable thing is that an empty ground floor does not mean the original article was empty. Even a two-line news brief would yield at least one entity and one information point. Total emptiness means the text never entered the system at all: the file failed to load, or it was image-only and optical character recognition returned nothing, or the page renders its body client-side so the scraper only ever saw a skeleton.

I ran into exactly this failure back when I was an editor in Shanghai. We had an automated tool pulling from roughly forty sources. Over six months it silently skipped every sports page from one large provider, simply because that provider changed its content management system and moved to JavaScript rendering. No alert fired. The dashboard stayed green. We only found out when a reporter asked why we were always missing coverage of one particular tournament.

Six months. Nobody noticed, because the number of articles published per day never changed. We were not short of content. We were short of one kind of content, and that shortage was masked by the surplus of another.

That is the first lesson, and the most frequently ignored one: a system returning empty does not raise an alarm. It goes quiet. And quiet, in a newsroom, is always filled by something else.

Layer two: what gets identified

Suppose the ground floor works. Text gets in. Information points exist. The next question is: who reads them?

This is where I want to slow down, because it is discussed less often. Over the past decade, sports data analysis has undergone a quiet coup. The analyst left the stands and walked into the dressing room. He no longer stands outside observing and narrating. He stands inside, spreadsheet in hand, telling the coach that the right winger is covering 8% less ground than last season.

That coup produced a great deal of good. It also produced a problem: the analyst's conclusions are frequently detached from the actual rhythm of the match. A model can say Team A should switch to attacking down both flanks. The model does not know that Team A's training-house internet was unstable for three weeks, or that their jungler just went through a breakup.

In esports, that gap is wider than in football, because public data is far thinner. Football has expected goals, passes allowed per defensive action, distance covered, penalty-area entries, player valuation models with dozens of variables. Professional esports, outside a handful of basic indicators such as win rate, KDA, creep score per minute and pick-ban rate, keeps most genuinely tactical data inside the coaching staff and never releases it.

Which means: an esports analyst who is honest must admit he is viewing a building through a keyhole. Most writers do not. They describe the whole building.

Layer three: what gets interpreted

I have had a mandatory rule since 2026, after the Shanghai derby night I have told many times and will tell again here because it still holds: every article must carry at least three different indicators before any judgement is offered.

That night, Shanghai Shenhua beat Shanghai SIPG 2–1. SIPG took 20 shots and generated 2.8 expected goals; Shenhua generated 0.9. My editor asked for a piece on fighting spirit. I refused and wrote that the win was luck. Fans burned my inbox for four days. The analysts read it.

What I did not say in that piece, and what I only understood later: three indicators do not save an article. Three indicators only guarantee that the article has an anchor. An anchor does not guarantee the boat is pointing the right way.

So at the interpretation layer there are three traps that Vietnamese esports writers fall into more often than anyone else — not because they are worse, but because they work in an ecosystem with too little public data and too much pressure to have an opinion.

Trap one: turning one match into a trend. A team loses three straight group-stage matches, and people write about a "form crisis". Three matches is three data points. With a baseline win rate around 50%, the probability of losing three in a row is roughly 12.5%. In a twelve-team tournament, on average one and a half teams will land in that state with nothing unusual happening. Writing about it as a tactical event is reading noise as signal.

Trap two: using pick rate as tactical evidence. Pick and ban rates are data about coach behaviour, not about champion strength. When a champion is picked in 90% of games, that may mean the champion is strong, or it may mean every team believes the champion is strong. Those two explanations lead to opposite conclusions, and only head-to-head win data can distinguish them.

Trap three: using outcomes to reverse-engineer decisions. A team makes a substitution at minute 30 and wins. The next day's article calls it a masterstroke. But if the shot at minute 88 had hit the post instead of the net, the same decision would be called reckless. I walked straight into this trap in the Euro 2026 semi-final, when I stated on radio that Denmark would beat England because they covered 6.4 more kilometres per match and took seven more shots. Denmark lost 1–2 after extra time. The indicator I ignored was not in the spreadsheet: squad depth, and the capacity of a substitute to change a game.

These three traps are not moral failures. They are methodological failures. But when they repeat at the scale of an industry, they become something else.

Anatomy of an empty analysis

Back to that report. I spent four hours reading it backwards, not to find content but to understand its structure. And the structure turned out to be very familiar.

It had nine sections. Patch and meta analysis. Tournament system analysis. Team and player analysis. Regional landscape. Club finance. Rules and governance. Risk profile. Public narrative and expectation. Industry transmission.

Nine sections. A handsome frame. In a normal report, those nine sections would be filled with team names, player names, patch numbers, revenue figures, sanctions, indicators.

That report did not fill them. In every section, instead of filling, it wrote one line: insufficient information. And at the end it added a passage I read three times, because it is something I rarely see in an analytical document: a warning about the writer himself.

The gist: the greatest risk in this analysis is not competitive, financial or personnel risk. The greatest risk is fabrication risk. When the input is empty, the analyst's natural reflex is to fill the frame with industry-plausible content — a guessed game title, a generic patch narrative, an invented transfer. And what gets produced will be indistinguishable in tone from real analysis. It will be consumed as fact.

I sat still for a while after reading that line.

Because that is precisely what is happening in Vietnamese esports, every day, at industrial scale.

The VCS and the thinness of public data

I follow the VCS from Shanghai through livestreams, usually on a time offset, usually at two in the morning. Based on my experience tracking matches across many seasons, there is a feature of the Vietnamese ecosystem I have not seen fully described anywhere: the VCS has one of the lowest signal-to-noise ratios among the major regional leagues.

What does that mean? It means the number of broadcast matches is high, but the amount of verifiable data is low. Scoreboards, pick-ban rates, creep score per minute — those exist. But the things that separate a champion from a fourth-place team — roster structure, communication allocation, coaching quality, operating budget — are almost entirely absent from public view.

That gap gets filled by two kinds of content. The first is impressionistic analysis: the writer narrates the match using adjectives. The second is pseudo-quantitative analysis: the writer uses numbers, but the numbers are selected to confirm a conclusion already reached.

Neither is analysis. The first has no anchor. The second has an anchor dropped after the boat has already docked.

I am not saying every Vietnamese esports writer does this. I have read genuinely excellent pieces by domestic authors, pieces I wish I had written. The problem is not individual. The problem is the incentive structure: a piece with data and a piece without data receive the same advertising, the same shares, the same editor satisfaction. Under those conditions, writing without numbers is cheaper. And cheap always wins.

That is why I say this is not a story about a broken system. It is a story about a price.

The economics of fabrication

I want to frame this in numbers, because that is the only way I know.

Suppose an esports outlet needs to publish 40 pieces a week. The marginal cost of an impressionistic analysis — no data collection, no verification, no interviews — is about two labour hours. The marginal cost of an analysis with real data is about twelve to sixteen labour hours, including collection, cleaning, cross-checking and writing the limitations section.

A gap of roughly seven times. On the same editorial budget, the economically rational choice is to produce seven impressionistic pieces instead of one data-driven piece.

But one variable changes the equation, and it is the variable most newsrooms ignore: the cost of correction.

A wrong impressionistic piece does not generate an immediate correction cost. It generates an accumulating one. Readers do not abandon a site over a single bad piece. They abandon it after two years, when they realise that none of those pieces ever helped them predict anything.

That is why sports data platforms survive. Not because they are better written. Because they can be wrong and know where they were wrong.

I paid for this lesson with a contract. In 2026, my research on empty stadiums was required to carry an optimistic closing paragraph about recovery. I refused. I lost the freelance arrangement with that outlet. The research was later cited by several Bundesliga coaches. I do not consider that a victory. I consider it a transaction with a price.

Every crowd is wrong. The only thing that is not wrong is probability. But probability only pays the people willing to count.

The counterintuitive angle: more data does not mean better analysis

At this point I have to argue against myself, because otherwise this piece is just a moral appeal, and moral appeals are the cheapest thing in the world.

My hypothesis above can be summarised as: thin data produces fabricated analysis. Therefore the solution is more data.

That hypothesis fails at one important point.

I have watched the platforms with the thickest data in the industry produce the worst analysis. Not because they lack numbers. Because they have too many numbers and no selection model. When you have four hundred indicators, you can prove any thesis by picking the twelve that fit. That is a more sophisticated form of fabrication, and far harder to detect, because it comes with numbers.

I call it evidence-backed fabrication. It is the most dangerous kind.

Second, more data does not solve the causality problem. In esports we constantly see indicator pairs move together and rush to assign causation. Winning teams usually have more kills. True, but kills are an outcome, not a cause. Winning teams usually control major objectives better. True, but objective control is a consequence of map pressure, which in turn comes from lane advantage.

I learned to write this section after the Euro stumble. Every piece I write now ends with a section called "Where might the assumptions be wrong?". That is not a humility ritual. It is a quality control instrument.

Third, and this is the point I most want to press: in esports, what is missing is not data. It is falsifiable data.

A good indicator is one that can prove me wrong. Win rate can do that. Ranking points can do that. But "a dip in form" cannot. "A shifting meta" cannot. "Roster depth" cannot. Those concepts cannot be wrong, and therefore they cannot be right.

From the Bundesliga to Worlds, I look for the same thing: a repeatable truth. Not a good story. A repeatable truth.

What actually exists in Vietnamese esports data

For this piece not to fall into the very trap it criticises, I have to list what I actually know and clearly mark what I do not.

What I know for certain. GAM Esports is Vietnam's most successful League of Legends organisation by VCS titles, and has repeatedly represented the region at the World Championship. Đỗ Duy Khánh, competing as Levi, is the jungler long associated with that organisation and one of the most internationally recognised Vietnamese players. Trần Duy Sang, competing as Kiaya, has been GAM's top laner across several recent seasons. These facts are verifiable through public tournament records.

At the international level, Faker — Lee Sang-hyeok — won the 2026 World Championship in Seoul with T1 and a fifth title in 2026 in London, where T1 defeated Bilibili Gaming 3–2. Those are traceable facts.

On transfers, Luka Perković, competing as Perkz, moved from G2 Esports to Cloud9 in late 2026, reported by international media at a buyout of roughly five million US dollars, then regarded as the highest figure recorded in League of Legends. I cite that number as a sourced fact, not as my own estimate.

What I know I do not know. I have no data on VCS team operating budgets. I have no salary structure data. I have no rights-fee or sponsorship revenue figures. I have no data on practice hours, coaching structure, or the rate at which analytical data feeds into draft decisions.

That gap is not my failing. It is a feature of the ecosystem. But what I can do — and what most writers do not — is state that it exists.

An honest analysis of the VCS under current data conditions would look like that report from the other night: full of N/A lines and a small section of verifiable claims. That is not a good analysis. But it is a correct one.

When sanctions arrive and nobody records them

There is one area where the data gap does the most damage, and I want to address it carefully.

Esports betting is eroding competitive integrity faster than traditional sport, and the main reason is regulatory lag. Football has had a betting monitoring system for decades, with independent bodies and public investigation processes. In esports, most of that system has been built within the last five years, and it has been built by the publisher itself — the party that both sets the rules and holds a direct commercial interest in outcomes.

The VCS events of 2026, with sanctions relating to match-fixing, are a case I followed. I will not name individuals here, because I do not hold the full text of the decisions and I do not want a piece about honesty to become a piece that is not honest.

But there is one observable point, and it matters more than the identity of those sanctioned: public reaction focused on individuals, not on structure. The story told was a story about bad people. The story not told was a story about a system in which mid-tier player salaries sit far below living costs, where an offer to fix one match can equal several months of income, and where the probability of detection depends on whether somebody happens to look at the data.

I say this not to excuse. I say it because a wrong diagnosis produces a useless remedy. And here, the remedy being applied is heavier punishment. Punishment only deters when the probability of detection is high enough. Detection probability depends on the quality of monitoring data. And monitoring data quality in Southeast Asian esports remains very low.

That is why I say the data problem is not a technical problem. It is a governance problem.

What happens when an empty analysis gets published

I want to return to the opening question through a concrete scenario, because abstraction is easy and specificity is hard.

An editor receives the empty report. He has a four-hour deadline. He has three options.

Option one: do not publish. The cost is a content gap. Nobody punishes him for it, but nobody praises him either.

Option two: publish a short note saying the source could not be verified. This is the journalistically correct option. It is also the one that gets the fewest reads.

Option three: write an analysis based on general industry understanding. He knows how esports works. He knows what problems teams typically face. He can write a perfectly coherent piece without a single verified fact.

Of the three, option three is the only one that creates no immediate career risk. That is why it gets chosen, every day, everywhere.

What that editor does not see is the second-order cost. Every such piece teaches readers a habit: that analysis is a literary genre, not a tool. And once readers are used to that, they can no longer tell a data-backed piece from a data-free one. At that point, the writer who has data loses his only competitive advantage.

That is the price. Not the price of one wrong article. The price of a market that has forgotten how to read numbers.

Where might the assumptions be wrong?

This section is mandatory, and for this piece it runs longer than usual, because I am writing about a subject on which I do not hold complete data.

Assumption one: I assume the empty report represents a widespread phenomenon. In reality I have a sample of one. A single case does not establish a rule. If this was an isolated incident caused by one specific system failure, my entire "economics of fabrication" argument loses its empirical anchor and becomes educated speculation.

Assumption two: I assume data-driven analysis costs seven times more to produce. That figure is my own estimate from personal experience, not from a survey. If the real ratio is three times rather than seven, my conclusion still points the same way but is substantially weaker in magnitude.

Assumption three: I assume readers want data. This is the assumption I am least sure of. It is possible that most esports readers do not want data; they want stories, and data is decoration that confers a sense of authority. If so, the market is functioning correctly, and the data-driven writer is selling a product nobody ordered.

Assumption four: I assume the absence of public financial data in the VCS is a defect. It may be a deliberate choice, and it may protect organisations from external pressure during a period when the industry is not yet stable. I have no evidence to rule that out.

Assumption five, and most important: I may be confusing cause with symptom. It is possible that thin data is not the cause of poor analysis but the symptom of something deeper — for instance, an ecosystem in which analysis has no commercial function, and therefore nobody pays for it to be done well. If so, every methodological remedy is meaningless.

I record these five assumptions not to retreat. I record them so the reader knows exactly where my argument can break.

Signals for the next cycle

If I had to bet on one thing over the next twelve months, I would not bet on a team, a player or a title. I would bet on data.

Specifically: the number of verifiable public esports data sources for Southeast Asia will increase, and it will increase not because of reader demand but because of sponsor demand. Sponsors are the only party in this chain with a financial motive to demand figures. They need to know what they are paying for. When they start asking, answers will have to be built, and once built, they will exist publicly.

I watched this cycle in the Bundesliga. It took about seven years, from the point shirt sponsors began demanding detailed reach reporting to the point advanced metrics became ordinary television language. Esports will move faster, because the data infrastructure already exists and younger audiences accept numbers more readily than traditional ones.

But there is an accompanying risk I want to record in advance: when data becomes currency, it will be faked. Not crudely — nobody invents an indicator that does not exist. They will select, splice, redefine, and compare incomparable periods. That is the hardest form of fakery to detect, and it will be the problem of 2028, not 2026.

They told me I was causing trouble. I was only reading the ending a few months early.

Transfers are a fertile gamble, but I count cards before I place a bet. And in this gamble, the only card I genuinely hold is the right to say I do not know.

On derby night in Shanghai, I chose the numbers over the entire city. Tonight, in Shanghai, I choose an empty spreadsheet over a full article.

With no crowd, football changed its nature. I found that out — and was rejected. With no data, esports changes its nature in a different way, and this time I do not intend to let anyone reject that before it is recorded.

The question I leave behind, and it is one I genuinely want asked rather than a rhetorical flourish: when a Vietnamese esports content platform publishes its next analytical piece, does anyone in that newsroom know exactly how many verifiable facts the piece rests on? If the answer is no, then everything else is built on sand — and sand does not hold any shape for long.