A Nine-Dimension Analysis With Zero Data Points

Last week I opened an analysis file running close to four thousand words. It contained a patch-impact table, a format-structure table, a roster assessment table, a six-row risk matrix, a three-tier industry transmission map, and twelve carefully defined technical terms. Every content cell carried the same line: “N/A — insufficient information.” Tournament name: N/A. Team name: N/A. Player: N/A. Patch number: N/A. The list of input information points was empty in the literal sense, not a single item. The only field in the entire file holding real data was a two-word label: esports.

People call me a “number freak”; I take that as a compliment. But the label carries an expectation too: that I will fill those empty cells with something — a guess, a hunch, a sentence beginning “in my view.” I sat with that file for a long while, and decided not to fill them.

That was the hardest decision of my week, and it is the subject of this piece.

The market wants answers before the data arrives

The Vietnamese esports calendar barely has an off-season. The Spring VCS closes and the Summer split goes live, the Arena of Valor league runs in parallel, Free Fire and PUBG Mobile events slot in between, then come the regional stages and the SEA Games. Every match generates two content products: a preview published within 24 hours of the start, and a retrospective within a few hours of the final whistle. Fall behind that rhythm and you lose the readership, and readership is the entire revenue line.

That pressure produces a familiar paradox: the faster you go, the thinner the input. A preview written at 11 p.m. for a 5 p.m. match the next day usually has three things in hand — head-to-head history, form over the last five matches, and some story about the locker room. All three are second-order data: already filtered through someone else's telling. Nobody in that production chain goes back to the primary source to ask one simple question: what does this number actually measure.

I learned that question somewhere with no connection to esports. In 2026 I was 19, a statistics student in Nha Trang, writing a V-League blog from a rented room. In round eight that season, one team held 61% possession and took 15 shots but generated only 0.8 expected goals; their opponents took 3 shots, posted 0.6 xG, and the match ended 1-1. I went back through every phase of play to understand how 15 shots produced less than 3. The answer sat in shot location and running distance: possession does not create truth, it creates an impression. Manual charting took four hours per match. That was my first standardisation process, and it taught me that every claim must have a traceable path back to primary data.

In 2026 I applied that process to the World Cup. Before the tournament I published a warning about Germany: their average PPDA had risen from 8.1 to 11.6 in qualifying, high-speed running distance had fallen by nearly 18%, most visibly in midfield. My conclusion was that Germany would go out in the group stage. The forums called me a “number freak,” and Germany finished bottom of Group F.

In 2026, when football returned to empty stadiums, I collected 64 Bundesliga matches. Home win rate fell from 42.7% to 31.3%; home xG dropped 0.19; away teams' PPDA improved by 0.8. I wrote that home advantage was really a noise advantage rather than a pitch advantage. A sports data company in Ho Chi Minh City read it and hired me as an analyst.

In 2026 I was tasked with building a World Cup model, standardising 68 teams into 12 metric groups. Before the knockouts I identified Morocco as a special case: they averaged only 28% possession but forced opponents down 0.35 xG per match, and goalkeeper Bounou posted PSxG overperformance of +2.4. At the same time, Argentina were the only side keeping PPDA below 8.0 in every match. I was criticised for removing Brazil from the contender list. Both teams I picked reached the final.

I wrote a blog from a rented room in Nha Trang; now probability takes me everywhere. But what I carried out of that room was not a set of numbers, it was a habit: before I write anything, I must be able to point to which data is holding the sentence up. When that path breaks, the conclusion is not “weak.” It does not exist.

Every analytical dimension is a list of preconditions

A serious esports analysis needs nine dimensions. I did not invent them to make the story sound grand; I have them because I got all nine wrong at some point. What matters is that each dimension has a minimum precondition list, and when the preconditions are unmet, that dimension does not yield a weak conclusion — it yields none.

Patch and meta. To say anything about an update you need four things: the game title, the patch identifier, the list of adjusted champions or weapons or maps, and win-rate plus pick-ban deltas between two versions. Without the game title the other three are meaningless. Three major titles run on three different rhythms: League of Legends on a two-week cycle, Arena of Valor by season, Free Fire by OB builds. Those rhythms produce three different kinds of crisis. A stat adjustment within roughly 5% usually causes mild disruption with a short adaptation window; a mechanic rework or map rotation can invert an entire hierarchy. I cannot say which kind the update in that file was, because the file has no patch number.

There is a specific risk I always screen for before an event: a patch targeting a dominant playstyle. But that claim is only valid when a team's win condition has been stable across several consecutive patches. Without that history there is no allegation, only inference.

Format. BO1, BO3 and BO5 differ in variance, and variance is the single largest lever on upset probability. A single-elimination bracket inflates the apparent strength of weaker teams; a double-elimination bracket with BO5 rewards adaptation and depth. The VCS plays BO3 in the group stage and BO5 in playoffs, which means a team with a narrow champion pool can survive groups and die in playoffs. That is a claim I will write, because it rests on a published format. But with that file, the tournament name is blank, so even that sentence has to be labelled an assumption.

There is a less-noticed interaction: schedule density multiplied by patch recency. A compressed calendar immediately after an update rewards teams with strong analytical coaching rather than teams with stronger individual mechanics. Testing it requires both the calendar and the patch release date. Neither exists.

Roster and players. Replacing three or more starters is a rebuild, not a reinforcement; the synergy cost is higher and the adaptation window longer. The age curve is role-specific: reaction-heavy roles such as mid lane and AD carry in MOBA, or entry in FPS, decline earlier than shot-calling and support roles. Levi (Do Duy Khanh) and Optimus (Tran Van Cuong) are examples of players who held a stable win condition across multiple seasons — which makes them the group most exposed to a targeted patch, and also the group most easily misjudged by a writer with no champion-pool data.

Region. Regional standing is title-specific and must never be generalised. The same country can sit at completely different levels in League of Legends, Arena of Valor and Free Fire. Without the game title, any regional claim is methodologically invalid, even when it sounds entirely plausible.

Club finance. A salary-to-revenue ratio above 80% is the norm that makes most esports clubs structurally loss-making. Sponsor concentration — one sponsor exceeding half of revenue — is a standard red flag. Parent-company contagion is a recognised systemic vector. But without a revenue breakdown and without naming the capital backer, none of it is measurable.

Rules and governance. Esports governance is publisher-centric: the publisher writes the rules, owns the IP, holds commercial stakes, and there is no independent arbitration body. That structure shapes every punishment scenario. But compliance analysis only means something against a specific alleged act — match-fixing, account boosting, approaching contracted players, dual contracts. No act, no analysis.

Risk profile. A risk matrix with no subject has no row to assess. This is where I want to slow down: no red flags here means no data, absolutely not no risk. Those two sentences are worlds apart, and in this profession, confusing them is the fastest way to lose money.

Narrative and expectation. A media story passes through four phases: budding, accelerating, climax, backlash. Shouting at the climax is planting the backlash yourself. The way to test whether a story is organic or manufactured is cross-channel consistency: mainstream press, vertical esports press, and short-video channels. Too much divergence means someone built the story.

Industry transmission. The flow runs from publisher to clubs and events, then to sponsors and derivative markets. Each publisher operates on a different cadence, so each transmission path has a different lag. Without a publisher name there is no path to draw.

Nine dimensions. None of them runnable. The match is over, but the data is still there — except this time it never arrived.

The fingerprint of a failed extraction

One detail in the file stopped me longer than the missing data itself. The entities field, which should hold the list of teams and players, contained an instruction: “identify from the information points above.” Meanwhile the information points above were empty. That is a self-referential loop, and it guarantees an empty return every time.

Put together, it forms a clear signature: the domain label correctly populated, title and source left as N/A, article type left “unclassified,” an empty information-point list, and the entities field still carrying the template's own instruction text. Those four signals travelling together mean the extraction step never ran — not that the source article had no content. The difference between those two possibilities is the entire problem: one is a process fault fixable within the same cycle, the other is a genuinely empty article.

What worries me more is silent propagation. If an empty file like that is pushed into an automated chain without a null-check, the system will generate fluent, grammatical, terminologically correct analysis that is entirely fabricated. In Vietnamese esports content I have read plenty of previews carrying “form,” “head-to-head record” and “psychological edge” where not one clause was verifiable. They were not wrong for lacking numbers. They were wrong for lacking a subject.

The most useful thing to take from that file is a data-recovery ladder, ordered by priority. P0 is two things: the game title, and at least one named entity — team, player, tournament, or publisher. Without those two, no analysis is valid regardless of how elegant the model is. P1 covers the patch identifier, the tournament name with format and team count, and the source name. P2 covers publication date, regions, and any monetary figure. P3 is the alleged rule violation, if any.

The ladder sounds dry, but it is a working tool. Every time I receive a file I run the four levels in about three minutes. If P0 fails, I stop. Not because I do not want to write, but because writing on is a move from analysis into fiction.

Fluency is rewarded; verification is not

This is the part that irritates me most while writing this, and I will say it plainly.

The market rewards fluency. A piece with decisive clauses and crisp assertions gets shared far more than one carrying a footnote that says “input insufficient.” Readers are not wrong to react that way; they are consuming a media product, not a technical report. But the consequence is that writers have a commercial incentive to fill the blanks, and that incentive operates quietly, without anyone issuing an order.

I call it the analytical ego trap. When a nine-dimension scaffold is already built in front of you, the pressure to fill it is enormous — because leaving it blank looks like incompetence. Our profession has an under-discussed failure mode: not a wrong prediction, but a process with no gate. A wrong prediction costs one match. A process with no gate costs a whole season, because it produces hundreds of wrong predictions at once and nobody knows which one came first.

There is another point the crowd usually misses when it hears “N/A.” In ten years of watching, I have found the costliest decisions come not from a counter-intuitive conclusion but from treating the absence of a bad signal as a good signal. No red flag is not a green light. It only means nobody has turned the light on to look.

I want to add one more thing about fans, because this is where data writers like me often push ourselves into an unnecessary confrontation. When a Vietnamese team loses a match it should have won, the crowd has every right to be angry. Their emotion is a variable to be explained, not an error term to be removed from the model. My job is to measure what that anger correlates with — a specific play, a draft decision, or just the memory of the previous loss. Correlation is not causation, which is exactly why you measure instead of pronouncing.

A data-completeness label

If I had to carry exactly one thing out of this story, it would be a small proposal: every analysis should carry one line stating what its input contained and what it lacked. Three lines is enough. Which source. Which timestamp. And which part is an assumption. The cost is close to zero, the value is large, because it hands the power of verification back to the reader instead of demanding belief.

The new season will keep producing matches where the data says one thing and emotion says another. There will be patches that turn a champion pool from an asset into a liability within two weeks. There will be teams that win groups with a brand of esports nobody can model, then get eliminated in playoffs by the very thing that carried them there. I will sit down with each match, with a file that may be empty, and with a habit that does not change.

An empty stadium does not need spectators; it needs an analyst willing to look. And sometimes the only thing visible in the frame is a blank cell nobody has filled in.