Nine Layers of Esports Data: The Analyst's Discipline When the Spreadsheet Returns Empty Cells

Minute 31 of game five. The team leading by 4,800 gold walks into the dragon pit without vision. One engage, three bodies on the floor, Baron to the opponent, and ninety seconds later the game ends in a way that anyone watching the highlight reel would describe as the other team simply playing better. In my spreadsheet, that moment occupies exactly one row: lost fight, minus three kills, Baron lost, 4,800 gold swing. I rewatched that game four times over two days, not to find who was at fault, but to answer a much narrower question: if that row were all I had, would I dare write that the losing team lost because of vision?

The answer is no. A lost fight can come from vision, from a mid lane that was shoved ten seconds early, from a summoner spell burned at the wrong moment in minute 28, or simply from a composition that is stronger in exactly that window. The data row I had recorded the consequence. It did not record the cause, and it certainly did not record how many times I had promised myself I would never guess just to make an article look full.

This piece was born from an evening when every cell in my analytical framework returned the same phrase: insufficient information. I sat there, holding an empty dataset, and realised my profession carries two temptations. The first is misreading data. The second is more dangerous: writing beautifully about a dataset that does not exist.

An evening when every cell returned N/A

I work in Seoul, covering esports for the Korean market, but my eyes keep turning southeast, where Vietnamese teams play the matches international media only mention when there is an upset. Some time ago I built a nine-layer analysis frame for an article about a match I believed mattered: patch and meta, tournament format, roster and players, regional map, club finance, rules and governance, risk profile, public narrative, and industry transmission. Nine layers, each with a handful of mandatory indicators.

When I pasted the input data, the output was a long column of N/A from top to bottom. No tournament name. No patch number. No team. No player. No source. No publication date. Exactly one field was filled in properly: the domain label, which said this was esports.

There was a moment, and I mention it because I think many people in this trade will recognise themselves in it, when I wanted to fill in the blanks. I know enough about the Vietnamese esports context to write something that sounds entirely plausible: the publisher's two-week patch cadence, the stories of teams rebuilding after every transfer window, the story of a young player pushed onto the international stage too early. All of it true at a general level. All of it worthless for that specific match, because I had not a single line of data belonging to it.

That mistake years ago taught me that data never lies, only the reading of it does. But there is a variant of that mistake I only learned later: when there is no data at all, the worst reading is to read with your imagination and then present it in the voice of someone holding numbers.

Layer one — Patch and meta: strongest is not the same as cheapest

Every esports analysis begins with a question that looks simple: which playstyle does the current version reward. In League of Legends the patch cadence is roughly two weeks, and most adjustments sit within a band of plus or minus five percent champion power. Those changes rarely invert a competitive hierarchy; they reshuffle pick priority. A mechanical rework or a map rotation, by contrast, can flip an entire team hierarchy inside a single week of scrims.

The first thing I check is never a champion's win rate but its ban rate. Win rate is too heavily influenced by who is playing the champion; ban rate reflects the collective belief of an entire coaching ecosystem. When a champion is banned in more than forty percent of games but wins only around half, that is the signature of fear rather than strength. And fear can be exploited.

With Vietnamese teams I look at a narrower indicator: how many champions the mid laner and the jungler can play at an acceptable professional level. If that number is below four, the team has signed its own sentence in a best-of-five draft, where the opponent has enough bans to erase the fingerprints of a playstyle. I have watched this repeat across seasons: a team that looks excellent in the regular season, then walks into a deciding series and is strangled in the draft phase.

One technical detail viewers usually miss: the tournament server and the live server diverge. An international event may run on a build several weeks older than what players practise on. Teams prepare on live, arrive, and discover that interactions they thought were settled behave differently. This is the highest-frequency source of pre-event controversy, and it can only be verified if you know the exact tournament build number.

Meta is not what is strongest in the game; it is the cheapest available path to victory in that specific patch, with that specific roster. Esports does not need luck, it needs people who read the meta faster than the server does.

Layer two — Format: where mathematics decides who is allowed to dream

No factor moves upset probability more than format. I often put this number in front of younger colleagues, and it always buys a few seconds of silence.

Assume an underdog has a forty percent chance of winning a single game. In a best-of-one, the chance of winning the series is exactly forty percent. In a best-of-three it falls to roughly thirty-five point two percent. In a best-of-five it drops to about thirty-one point seven percent. Same team, same form, only the format changes, and the path forward shrinks by nearly a quarter.

That is why open tournaments with single-game qualifiers reliably produce surprise names nobody mentions three months later. They did not get better. They were simply placed inside a structure with high variance.

Brackets work the same way. A team drawn into a half containing the two strongest opponents burns tactical and physical energy earlier than an equally good team in the other half. Viewers see the scoreboard; analysts have to see the bracket, the schedule, and the number of rest days between series. At events involving intercontinental travel, time zone difference is a real variable, not an excuse.

I got this wrong once. At a regional event I predicted a team would go deep based on their last ten matches, forgetting they had to play three best-of-fives in six days while their opponent had four days of rest. They lost the third series, and the indicators I used were not wrong. What I missed simply lived outside the indicators.

Format is the only thing in esports analysis you can calculate with pen and paper, and it is the thing most often skipped.

Layer three — Roster: three changed names is a rebuild

There is an unwritten threshold in this industry: replacing three or more starters counts as a rebuild, not a reinforcement. Coordination costs spike, and the adaptation window typically runs one to two months of real competition.

But that number only means something if you know how long the previous lineup held together. A roster unchanged for three seasons has enough shared foundation to absorb one major change. A roster that has swapped players every window is simply continuing an established state.

For Vietnamese teams I look at three indicators before I look at standings. Gold difference at fifteen minutes, CS difference at fifteen, and the jungler's kill participation. Together they tell a fairly clear story: which team controls early tempo through lane pressure, which team lives on skirmishes, and which team is turtling and waiting for mistakes. A team with positive gold difference but negative CS difference across all three lanes is usually winning on individual skill rather than structure. That kind of winning looks beautiful in the regular season and breaks easily in a best-of-five against an opponent who reads the map.

Imports must be read through data rather than feeling. A Korean player joining a Southeast Asian league brings training discipline and fight comprehension, but also a language barrier during exactly the window when the team most needs clean communication. In shooters, the in-game leader is the centre of everything; in MOBAs, the thread between mid, jungle and support is where tempo is created. Replacing one node in that thread changes the entire information system.

Between the transfer numbers sits a story nobody writes in the report: the story of a player who was excellent on his old team and invisible on his new one, not because he declined, but because the new system did not need the thing he was best at.

Layer four — Regional map: a world slot is not a certificate

Regional strength is a game-dependent concept and must never be generalised. A region can sit at very different tiers in League of Legends, DOTA 2, CS2 and Valorant at the same moment. Vietnam is the clearest example of this stratification: strong in mobile titles, with a deep tradition and an enormous community in Arena of Valor, while on the PC international stage it has usually been placed in the group that must qualify to prove itself.

From the 2026 season, Vietnamese teams entered a restructured Asia-Pacific regional system. A structural change means historical head-to-head data loses part of its value: you cannot directly compare a team's record in the old system with the new one when the number of opponents, matches and scheduling density differ. An analyst must mark this explicitly, or the chain of evidence is corrupted at the first link.

The regional indicator I trust most is not title counts but how many academy players are promoted to the main roster and how many games they actually play. A healthy esports scene can produce replacements. A sick one buys them. A dying one can neither produce nor buy, and retains its best people with multi-year contracts that contain no development path.

Layer five — Finance: eighty percent of revenue goes to salaries

Esports clubs worldwide commonly run salary-to-revenue ratios above eighty percent. That figure makes the business model almost always loss-making on paper and turns every parent-company injection into a survival factor. So when I read a transfer story, I ask three questions: where the money comes from, who pays it, and over what period.

Sponsor concentration is a heavy warning sign. When one sponsor exceeds half of revenue, the club is no longer an independent sports organisation but a marketing channel with an expiry date. Parent-company contagion is a recognised systemic risk vector: when the money behind the club comes from real estate or a streaming platform, that sector's downturn transmits straight into the payroll.

In the Vietnamese market I watch a specific pattern: many large organisations are backed by media, retail or entertainment brands, and their value sits in audience size rather than ticket revenue or broadcast rights. That creates a paradox. The more famous the team, the easier the sponsorship, but also the stronger the pressure to hold results in order to hold attention. Short-term performance pressure usually leads to buying players instead of developing them, and that is the spiral smaller clubs cannot escape.

I once bet on the wrong dataset and received the right lesson. The dataset was the payroll of a club I assumed was stable. The numbers were not wrong; I was wrong because I did not know the money came from a sponsorship with a performance-linked renewal clause. When the team lost three in a row, the renewal did not happen, and my beautiful payroll became a liability.

Layer six — Rules and governance: the referee sits at the owner's table

Esports governance has a structural feature anyone analysing risk must remember: the publisher writes the rules, owns the intellectual property, and holds commercial stakes in the very competition it referees. There is no independent arbitration body equivalent to a sports court in football.

That does not make esports governance arbitrary. It means every ruling must be read against the interests of the party issuing it, and any comparison of sanction severity across regions is valid only when the same ruleset applies.

The usual violation categories include match-fixing, account boosting, approaching contracted players, dual contracts, and issues involving underage competitors. In Southeast Asia, competition bans for match-fixing have appeared repeatedly in recent years, and their impact extends well beyond the individual sanctioned: an entire roster loses a season, sponsors withdraw, and audience trust in the integrity of the league is damaged for years.

What I want to stress is this: a compliance analysis only means something when there is a specific alleged act. Without one, listing legal risks is a checklist, not a conclusion. And a checklist with no boxes ticked must never be presented as a clean bill of health.

Layer seven — Risk profile: an empty cell is not a clean certificate

This is the layer I believe matters most and is most often misread.

A risk profile has six categories: competitive, financial, personnel, rules, public opinion, and systemic. Each requires an identified subject. When no subject exists, the profile cannot be scored, and the only defensible statement is that it cannot be assessed.

In my trade there is one error that ends careers faster than any other: reading the phrase insufficient information as the phrase no problem found. Those two sentences are logically opposite, but on a neatly formatted report they look identical.

I learned this from something unrelated to esports. The cancelled Seoul derby of 2026 was a stress test for every prediction algorithm. My model at the time was built on ten matches from the opening weeks, and when the league stopped, the entire denominator vanished. Numbers I had treated as foundations became numbers with no source. The analysis I wrote then was rejected by my editors as too sensitive, and I kept it in my archive. A year later I reread it and found it was not technically wrong. It was wrong in presenting conditional reasoning as settled conclusion.

Since then, every judgement in my writing carries a confidence level. High, medium, low. And for low-confidence judgements, I state explicitly which data they depend on.

Layer eight — Narrative and expectation: the crowd does not read spreadsheets

A team can be performing exactly as the model predicts and still be described as being in crisis, purely because of storytelling. A media narrative typically passes through four stages: emerging, accelerating, peak, and backlash. Most of a team's damage comes not from the peak but from the backlash, when the people who once celebrated turn to hunting reasons to criticise.

The best test is cross-channel consistency. When mainstream media, specialist press and community channels tell the same story with the same intensity, the story is likely manufactured. When those channels diverge, you are watching a fact being contested.

The betting market is not wrong; it merely reflects a truth you have not yet seen. I track odds movement as an indicator of crowd expectation, and I always place it next to real data to find the gap. When expectation and reality diverge, there are two possibilities: the crowd is wrong, or I am missing information the crowd already has. Most of the time, the second is true.

I do not trust intuition; I trust numbers that speak after being asked the right question.

Layer nine — Industry transmission: the publisher's cadence is the heartbeat

Everything in esports flows to the publisher's rhythm. A two-week cadence from one major MOBA publisher, infrequent updates from a distribution platform, and a season-based model from an Asian technology conglomerate produce three fundamentally different ecosystems. Those three cannot be analysed with the same toolkit.

The flow runs from the upstream publisher, through the midstream of clubs, tournaments and broadcast platforms, down to downstream sponsorship, derivative products and mainstreaming. When upstream cuts budget, midstream feels it first through events being shrunk or outsourced. When midstream contracts, downstream loses its channel and withdraws money. The lag between the three layers is usually two to four quarters.

In Vietnam, the downstream layer has a notable feature: the penetration of mobile esports into daily life is far greater than that of PC esports. That creates an opportunity and a risk simultaneously. The opportunity is audience scale. The risk is that the commercial value of that scale is systematically undervalued relative to its popularity, because sponsors have not found a way to convert attention into revenue.

The contrarian angle: correlation is not causation, and fluency is not truth

There is a template I see almost weekly online: a team changes its jungler, wins three in a row, and a conclusion is drawn immediately that the change solved the problem. Three matches are three data points. A season is roughly forty. You cannot conclude a season from three matches, and you certainly cannot conclude causation from a short time series.

But the temptation does not lie in the wrong conclusion. It lies in how good the wrong conclusion can look. Fluent writing creates a feeling of certainty, and that feeling is what makes readers believe without checking the source.

That is why I built a rule for myself: if a piece cannot name its original data source, its publication date, and the boundary conditions of that data, it does not leave the draft folder.

For the same reason, I learned to write the sentence many in this trade avoid: there is not enough information to conclude. That is a professional sentence, not an evasive one. A good analyst knows exactly what they do not know, and can say so without losing credibility.

With the empty dataset I mentioned at the start, I considered writing a full analysis of a match I had no information about. I knew exactly how it would look: open with a moment, cite a few general regional indicators, close with a two-sided prediction that could never be fully wrong. It would be shared. It would collect a few compliments. And it would be a structured lie.

I did not write that piece. I wrote this one.

Every season is a ritual, and the analyst is merely the person who records the omens. But the recorder carries an obligation: when there are no omens, record that there are no omens, instead of inventing a beautiful one.

What to track next

Over the coming months I will watch four signals closely. First, the mid-season patch cadence, because that is when teams have the least adaptation time and when gaps in coaching quality show most clearly. Second, regional format structures, because any change to the maximum series length shifts upset probability in a calculable way. Third, the mid-season transfer window, where struggling clubs tend to decide emotionally and pay with next season's budget. Fourth, sponsorship cash flow, the earliest indicator that an organisation is preparing to contract or expand.

None of those four signals gives me a certain prediction of who will win. They only tell me where to place the question.

And perhaps that is the whole of this profession. We do not live on answers. We live on knowing which data we are missing, and on having the nerve to say so before anyone asks.