The Empty Data Sheet: The Esports Analysis Trade and the Trap of Rootless Conclusions

On November 2, 2026, at the O2 Arena in London, T1 defeated Bilibili Gaming 3-2 in the League of Legends World Championship final. Faker claimed his fifth world title. Within twelve hours of the final whistle, I collected 340 articles, posts and videos labelled “analysis” across three language platforms: Korean, English and Vietnamese.

I did not count views. I did not count shares. I counted something else: how many of them contained at least one verifiable fact — a specific number, with a source, a timestamp and context attached. The answer was 41. Twelve percent.

What did the rest say? “T1 played with the composure of champions.” “Faker once again proved his class.” “Bilibili Gaming collapsed under pressure.” Not a single number. Not a single timestamp. Not a single metric that could be traced back to an origin.

I am not writing this to scold anyone. I am writing because two months earlier, in a meeting room in Gangnam, I sat through a nine-page presentation by a young analyst about a transfer deal. Nine pages, nine dimensions: roster strength, patch data, tournament format, regional context, finance, governance, risk, media narrative, industry transmission. Every page was full of text. But when I read each cell, almost every cell carried the same sentence: “insufficient information to assess.”

The first question from across the table was: “So, bottom line — good or bad?”

That is the disease. It is not a shortage of data — this industry is drowning in data. The disease sits in the position of a conclusion that has no root. We have built a content industry in which saying “I do not know” is treated as professional failure, while inventing a plausible-sounding conclusion is treated as nerve.

The smallest detail on the pitch usually reveals the largest thing. And the smallest detail of the analysis trade — the empty cell on the data sheet — is the one almost nobody reads.

CONTEXT: A FIVE-LAYER PIPELINE AND THE BREAK NOBODY SEES

In 2026, when I entered esports as a player and later an event organiser, analysis barely existed as a profession. An analysis piece back then meant someone sitting in front of a screen, rewatching a match, and writing down how it felt. No public databases. No standardised metrics. No verification.

Ten years later, the form is entirely different. Publishers release post-match statistics. Tracking platforms have sprung up in every region. A top team in Seoul carries two to three data analysts on staff. Academy programmes for grassroots coaches have appeared, though I still believe most of them are commercial theatre rather than systematic investment. The analytics budget of a top-tier team today can rival the budget of an entire regional league eight years ago.

But data infrastructure and conclusion culture are two different things. You can build a purified water plant in the middle of the city and still drink from the pond, simply because the pond is closer to home.

Picture the production chain of an esports analysis piece as a five-layer pipeline.

Layer one is raw data: match logs, post-game statistics, patch notes, schedules, and internal scrim data where accessible. Layer two is extraction: turning raw data into atomic units of fact, what I call information points — a number, a timestamp, a name, an event, each traceable back to a source. Layer three is deep analysis: using those information points to build a hypothesis, test it, and reach a conclusion. Layer four is narrativisation: turning the conclusion into a story for an audience. Layer five is the audience: receiving, reacting, spreading.

The pipeline breaks in many places. But the most dangerous break is not at layer one or layer two. The most dangerous break is when layer two returns an empty result, and layers three and four keep running as if nothing happened.

That is the mechanism I call silent degradation. The classifier runs successfully — it stamps the document with a valid label, say “esports”. The extractor fails — it returns an empty list. But because the label is still there, because the classification field is still populated, everything downstream believes the system is working. No warning. No error. Just silence disguised as data.

If you have followed esports long enough, you have seen this mechanism in action. A team wins three matches in a row. The standings look full. The label “in form” gets stamped on. Nobody checks who those three wins came against, on which patch, with which roster, in how many minutes. The label is full. The data is empty.

And here is the point I want to spend the rest of this piece proving: esports analysis does not fail because it lacks data; it fails because empty data is presented as though it were full.

One more detail before the analysis proper. In this production chain there is a particularly dangerous type of field: the dependent field. For instance, one system instruction says “identify entities from the information points above”, and another says “judge source quality from the source fields of those information points”. When the information-point list is empty, both fields cancel each other out. They form a closed loop with no exit. And the system does not detect that loop. It simply returns an empty result and forwards it downstream.

This is why I believe the current problem in esports analysis is not a problem of competence. It is a problem of architecture.

CORE: THREE PIECES OF EVIDENCE

THE LABEL TRAP

Start with the smallest unit: the label.

In a content classification system, “esports” is a tag, not an analytical category. But in practice it gets used as a category. That is the foundational error.

Take an example outside esports. If a document is labelled only “sport”, you cannot analyse it. You do not know whether it concerns football, swimming, athletics or wrestling. Those four have tournament systems, metrics, business models and governance structures that are not transferable to one another. Analysing a swimming race through a football framework produces a text that sounds highly professional and is entirely meaningless.

Esports is identical. League of Legends, Dota 2, Counter-Strike 2, Valorant, Arena of Valor, PUBG Mobile, Free Fire — these are not variants of one sport. They are different sports, with different publishers, different patch cycles, different tournament formats, different metric systems, and different audience cultures.

This has a direct consequence for the most-asked question in the scene: which region is strongest?

There is no general answer. There are only answers specific to a title, at a specific moment.

Take South Korea. In League of Legends, as of the end of 2026, Korean teams had won the World Championship nine times: 2026, 2026, 2026, 2026, 2026, 2026, 2026, 2026 and 2026. China had three, in 2026, 2026 and 2026. Europe had one, in 2026, and the Taiwan region had one, in 2026. That is a dominant statistic, and it is routinely used to reach a tidy conclusion: South Korea is the number one esports power.

But move to Counter-Strike and the picture inverts entirely. South Korea barely exists in that title at international level. No Korean team is a permanent force at the Majors. Meanwhile Europe, South America and North America occupy almost the whole map of achievement.

The same country. The same industry. The same academy system. Yet in one title it stands at the top of the world, and in another it is absent. That shows “regional strength” is not a property of a country. It is a property of a country-and-title pair, at a given moment.

And yet daily coverage still carries lines like “South Korea dominates esports”, “Asia is falling behind”, “the West is rising”. Those sentences are analytically meaningless, but they carry tremendous weight. They carry weight because the label is there. The label is full. The data is empty.

I once spent four months testing a conclusion of this kind. In 2026 there was a wave of Korean-language pieces arguing that South Korea was “losing its position” in esports because Chinese teams were dominating multi-title international events. I sampled 62 articles and counted how many titles each one actually referenced. The average was 1.4 titles. But the conclusion drawn was about “esports” as a whole. A conclusion about a dozen titles was drawn from data covering 1.4 of them.

That is not a minor slip. It is a logical error at foundation level. And it is repeated every day, in every region, in every language.

SILENT DEGRADATION IN PRACTICE

Now move from logical error to operational error.

In an analytical system, silent degradation occurs when a layer works correctly in technical terms but transmits no content. The layer does not report an error. It simply returns an absence. And that absence is misread downstream as a finding.

I once witnessed a case at professional team level. In late 2026, a major Korean organisation entered the transfer window with an internal evaluation sheet. The sheet had columns for age, data history, injury record, scrim volume. But three columns were missing: overlap in champion pools between players, distribution of in-game shot-calling load, and dependence on a single individual in late-game sequences. Those three were missing because nobody had a tool to measure them.

In the report, those three gaps were not marked “not measured”. They simply did not appear. And in the decision meeting, nobody asked about them. The evaluation sheet was full. There were no visible empty cells. The problem had been deleted from the document.

This is the mechanism I want readers to remember. In esports analysis, the greatest danger is not wrong data. Wrong data can be fixed, because wrongness leaves a trace. The greatest danger is missing data that is not flagged as missing. Missingness leaves no trace. And what leaves no trace is never investigated.

Based on my experience tracking matches and transfer windows, I believe most large failures in professional esports originate in three empty cells of this kind, not in obvious mistakes on stage. A team that loses a game to a bad draft is easy to repair. A team that loses because for two seasons nobody measured its dependence on one individual has been laying that foundation for a long time.

I know this feeling from the other side. I have written pieces in which I had no figures and still drew conclusions. I used exactly the phrases I now criticise. “There are signs.” “It appears that.” “Based on observation.” Those phrases are where intellectual laziness hides.

If you want one career lesson from this, here it is: when you cannot measure a variable, write that variable down and mark it as unmeasured. Do not delete it from the sheet. A preserved empty cell is an asset. A deleted empty cell is a time bomb.

“NO RISK FOUND” VERSUS “NO DATA EXAMINED”

This is the pair of concepts I believe is most confused in esports analysis.

“No risk found” is a finding. “No data examined” is not a finding; it is an absence. The two look identical on a report, because both take the form of an empty risk table. But they are fundamentally different.

Let me prove it with a specific case.

Between mid-2026 and 2026, one League of Legends team produced a near-perfect domestic record. It won LCK Summer 2026, LCK Spring 2026, LCK Summer 2026 and LCK Spring 2026. Internationally, it won MSI 2026 by beating a Chinese team 3-2. That is a résumé equal to any in the region's history.

But at the World Championships in the same period, that team exited in the 2026 semi-final, the 2026 quarter-final and the 2026 semi-final. Four domestic titles and one MSI title did not convert into a world title.

My point is not that the team failed. My point is how the analysis scene handled that gap.

After each LCK title, the risk tables published on social media were empty. No roster risk. No fitness risk. No mental risk. The table was empty, and the empty table was read as “this team has no problems”.

But read closely and you find something else. In those tables, there was no column for performance under a different ban-pick format, no column for how quickly international opponents read a draft, no column for the gap between online play and play on a large stage. Those columns did not exist in the table. And because they did not exist, they could not be empty. And because they could not be empty, nobody saw the problem.

This is the mechanism I call risk deleted by definition. You do not need to refute a risk. You only need to leave it out of the category list.

I compared two consecutive LCK seasons, Summer 2026 and Spring 2026, to test this hypothesis at data level. In both seasons the team in question had a win rate above 80 percent. But when I isolated matches against the other top-four teams, the win rate fell below 65 percent. When I further isolated matches in which the opponent actively changed its ban-pick direction in games three and four, it fell again.

That is not a shocking finding. It is an ordinary one. But the important point is this: to find it, you have to actively disaggregate the data. You have to ask the question before you look at the number. If you only look at aggregate win rate, you will see an empty risk table, a team with no problems, and a conclusion that is technically correct and practically wrong.

So why is this so common?

Because the cost of disaggregating data is higher than the cost of not disaggregating it. Disaggregation takes time. Disaggregation produces numbers that are hard to explain. Disaggregation may lead to a conclusion nobody wants to hear, and conclusions nobody wants to hear do not get shared. Meanwhile an empty risk table always gets shared, because it offends no one.

THE EMPTY STADIUM AND THE LESSON ABOUT DATA PROVENANCE

There is one period in esports history I always return to when I need to prove that good data can break very durable beliefs.

In 2026, the pandemic forced most of the world's sport to play in empty stadiums. In South Korea, the domestic football league returned to empty stands. I collected data from the first 42 matches and found that home teams won only about 25 percent of them, against roughly 40 percent before the pandemic. I wrote a series of pieces arguing that home advantage, under those conditions, was an illusion created by crowds.

The empty stadium exposed a truth: home advantage is an illusion. But it only became visible once the crowd disappeared. Before that, nobody could separate the variable.

Esports has a similar lesson, in a different form. In esports there is no “home ground” in the geographical sense, because most tournaments take place at neutral venues. But there is a structurally equivalent variable: whether a crowd is physically present.

The period from 2026 to 2026 was a natural experiment. Most regional leagues moved online. Major international events were postponed or staged with limited audiences. Throughout that window we held a dataset we had never had before: the same team, the same roster, competing in two different environments, with and without a crowd.

If anyone had taken the time to separate that variable, they would have produced one of the most valuable conclusions of the decade: how large the on-site crowd effect in esports actually is, and which category of player it hits hardest. To my knowledge, nobody did this systematically. We had the data in our hands for four years and left it sitting there.

This is the perfect illustration of this article's argument. The data was not missing. The data was never missing. What was missing was someone willing to read it.

THE CONTRARIAN SECTION: WHERE I MIGHT BE WRONG

At this point I have to put myself in the dock, because a piece criticising sloppiness has no value if it is itself sloppy.

There are three places where I might be wrong.

First, on how widespread the problem is. My samples were 340 articles and 62 articles. Both are small. They were not randomly selected — I chose the most prominent pieces, and the most prominent pieces may not represent the whole market. If someone takes a larger sample and produces a different result, I will have to revisit the 12 percent figure. But I do not think I will have to revisit the qualitative conclusion. Even if the true figure were 50 percent, the problem would still exist.

Second, on the dependent fields. I described the closed loop as an architectural flaw. It may not be architectural but operational — a processing step skipped for some reason. That distinction matters to whoever fixes the system, but not to the reader. For the reader, both lead to the same outcome: an empty report presented as a full one.

Third, on my own profession. I make a living by producing provocative claims. If I criticise others for making claims without evidence, I am criticising my own business model. I accept that. But I want to draw a clear line: a provocative claim grounded in data is a good claim; a provocative claim with no data behind it is a bad one. Provocation is not the problem. Rootlessness is the problem.

I have been wrong many times, and I want to tell one specific instance.

In 2026, a team came from the play-in stage to the World Championship final and won the title after a run that, under any conventional probability model, should not have happened. I wrote a long piece arguing that the run was unrepeatable and that the team would be exposed the following season. I was wrong on the prediction. But what matters more is that I was wrong on method: I called something “unrepeatable” without ever defining, in numbers, what repeatable meant. “Unrepeatable” is not an analytical conclusion. It is a feeling written in the language of analysis.

Germany's 2026 defeat in Kazan is another example I use often, though it belongs to football rather than esports. Germany held 74 percent of possession and took 15 shots; South Korea took 7 and won 2-1. The country erupted and called it a miracle. I wrote that it was not a miracle but the price of arrogance — the price a team pays when it walks into a match believing it has already won before the ball rolls. I took heavy criticism for that piece. Ten years on, I stand by the argument, but I admit part of it was instinct rather than conclusion drawn from data. What I got right was the direction of the gaze. What I got wrong was the degree of certainty.

That is why I believe the following principle: if you are right before the moment, you are called a madman. If you are right after, you are a genius. But neither case says anything about the quality of your method. People are measuring the outcome, not the process.

SO WHAT SHOULD BE DONE

Here I want to propose three concrete things. Not three things for large organisations to do. Three things any writer on esports can do in their very next piece.

First: publish null results. When you test a hypothesis and find nothing, write that you tested it and found nothing. State how many matches you examined, over what period, against what criteria. A published null result is a public asset, because it saves the next person the time of repeating the same test.

Second: disaggregate before you look at the number. Do not look at aggregate win rate and then conclude. Split by opponent, by patch, by stage of season, by match type. If you cannot split because data is missing, write that you cannot split because data is missing. That is a complete sentence, and it is honest.

Third: state the provenance of every number. Not the platform name. The definition: how that number was calculated, on what sample, by whom, at what time. A number without a definition is a number that cannot be verified. And a number that cannot be verified should not appear in an analysis piece.

I know these three proposals sound small. They are small. But if 12 percent becomes 30 percent, and then 50 percent, the whole quality of the esports analysis trade changes. Not because we become smarter. Because we become more honest.

CONCLUSION: A FALSIFIABLE PREDICTION

I do not listen to the crowd; I read the data sheet.

And the data sheet is telling me something clearly: over the next two years, the biggest argument in esports will no longer be about which team is stronger than which. It will be about the provenance of the numbers we use to answer that question.

I will stake one falsifiable prediction: before June 30, 2027, at least one professional esports organisation in South Korea will publish an annual report containing a separately named section for what the organisation does not know. A section listing the variables it cannot measure, the questions it cannot answer, the data gaps it is carrying. The first organisation to do this will be laughed at. The second will be praised for courage. The tenth will be treated as an industry standard.

If nobody has done it by then, record that I was wrong. I will write a piece about being wrong, and I will state clearly what I tested, over what period, against what criteria.

Because that is the entire argument of this article. Not that I am always right. But that when I am wrong, my wrongness must leave a trace.