Nine Layers of Esports Data: A Polished Template and the Empty Cells Nobody Checks

A report of eleven pages landed in my inbox at 23:40 Seoul time. It had a title, a table of contents, neatly ruled tables, and all nine analytical layers that almost every esports data desk now uses to assess a season: patch and meta, tournament format, roster and players, regional landscape, club finance, rules and governance, risk profile, public narrative, and industry transmission.

I read it end to end in fourteen minutes. All nine layers returned the same result: insufficient information. No game title. No patch number. No tournament name. No team. No player. No timestamp.

Emptiness did not stop me. Confidence did. The final line read: “The analysis framework is complete.” Forty tables. Nine layers. Not one data point.

In Seoul, people call that a report. To me, it is a test of whether the reader actually opens every cell.

I learned this trade through football, and I learned it by hand. In 2026, while still a middle-school student in Busan, I sat and counted every pass in a K League 2 match between Busan IPark and Seoul E-Land. I recorded 412 completed passes for Busan. The official statistics sheet published 389. I posted the comparison on a forum, it caused a small argument, and I kept raw data from nearly 50 matches to verify myself. Four hundred and twelve passes, and the official number was a polite lie. From then on I carried exactly one rule into esports: never accept a ready-made statistics table without asking how it was built.

Esports, however, runs on a different rhythm. A Korean season stretches almost year-round, split into stages, and each stage arrives with at least one update capable of reversing the priority order of an entire lane. The publisher ships updates on roughly a two-week cycle, which means a team competing across a season passes through more than twenty different competitive versions. A report that says “Team A is stronger than Team B” without naming the version, the stage, and the drafting rules is really only saying that the author watched two matches and wrote down a feeling.

That is why the nine-layer template exists. In principle it is a reasonable checklist. The problem is that each layer only has value when raw data is poured into it.

An update without a version number and a release date is not an update; it is a rumour with tables attached.

The patch and meta layer requires a minimum of four things: version number, release date, pick and ban rates for key champions, and how far those rates shifted from the previous version. Without those four, the sentence “the meta is shifting” cannot be verified.

I once reconstructed the pick-priority chain of a Korean stage by hand-recording every ban and pick across 41 consecutive games, then cross-checking against the organiser's published sheet. The largest discrepancies fell in the opening games of the stage, when teams were still experimenting. That means pick-ban data has a stable lag: for roughly 48 to 72 hours after a new version hits the competitive server, every priority ranking is still noisy. Every pass leaves an ink trail if you bother to follow it, and in esports that trail is pick-ban history at game level, not a summary table at tournament level.

I also tag the confidence of every metric I use, on three levels. Level one is data I counted myself from match records. Level two is published data with a clear, reproducible definition. Level three is published data with no definition attached. Level-three metrics are never allowed into a conclusion, even when they match my intuition. The rule costs time, but it is the only fence between analysis and guesswork.

There is one more variable most reports skip: fearless draft. When a champion used in one game cannot reappear later in the same series, the entire logic of “the number one priority pick” collapses. A team can win game one with its strongest composition and then lose game four because it has run out of options in one lane. Champion win rates at that point no longer measure champion strength; they measure the coaching staff's ability to allocate resources. Same table, two entirely different meanings, and most reports will not distinguish between them.

The second layer is tournament format, and it is the most underrated layer in the whole industry. Series length, upper and lower brackets, Swiss format, how many teams advance, the gaps between match days — all of these act directly on the final result. A team with a high game win rate but a low series win rate has a problem adjusting between games, not a problem with individual skill. If a report does not state series length and format, it is comparing two things that do not share a unit of measurement.

Schedule density is a direct consequence of format and also the most forgotten variable. A team playing three series in seven days has a completely different form curve from a team playing three series in fourteen days, even if the total number of games is identical. When a team suddenly loses its ability to control teamfights in the fourth game of a long match day, the cause usually sits in the calendar, not in the tactics.

The third layer is roster and players. Here I need to be blunt: published individual metrics in esports vary wildly in quality, and the problem lies in definition. Average creep score per minute only means something in relation to game length, role, and team strategy. A mid laner playing on a map-control team will post a lower creep score than a mid laner on an early-game team, and that says nothing about their level.

Based on my experience following matches in both football and esports, I always check context before reading any metric. In 2026, I calculated South Korea's PPDA against Germany at 9.8, below the tournament average, and concluded they were pressing proactively rather than defending passively. A PPDA of 9.8 is not defending — it is how a team declares war with a number. But that conclusion only held because I isolated the match context: the scoreline, the timing, and the fact that the team had to win. In esports, the equivalent context is the game state and the role assigned to each position.

Roster depth is the fourth variable reports tend to miss. A roster with six players rotating across two positions is fundamentally different from a fixed five. Depth is not measured by the number of substitutes, but by how many tactical options a coaching staff can deploy without changing players. When the schedule is dense, this becomes the deciding variable.

Physical risk also belongs to this layer, and it gets treated as a footnote. Wrist and shoulder injuries among professional players are not isolated cases; they are the consequence of repeated high-intensity motion. A roster without a contingency plan at its highest-load position is holding a time bomb, and that bomb appears in no power ranking.

The fourth layer is the regional landscape. South Korea, China, Europe, and North America have very different academy structures and internal competitive density. A region with a functioning lower division produces young players at a steadier rhythm than a region relying on open scouting. When assessing regional strength, I do not read the international results table; I read how many official matches a young player accumulates before reaching the main roster. International results are the outcome, academy throughput is the cause, and only the cause forecasts the next cycle.

There is another regional variable I once measured carefully in football and still see neglected in esports: the audience. In 2026, when stadiums stood empty, I analysed the May and June stretch of the German league and found that one club's home expected-goals differential fell from a clearly positive figure to a negative one. The crowd left the stands, and the home-ground equation lost its largest variable. In esports, when a tournament shifts from an on-site event with a live audience to online play on a neutral server, the equivalent variable also disappears. Home advantage is not atmosphere; it is a number capable of evaporating.

The fifth layer is club finance. This is the layer with the least public data and the most inference. Main revenue streams are sponsorship, publisher distributions, and commercial revenue. The largest cost is almost always the wage bill. From the 2026 season, the Korean league applied a salary cap mechanism with certain exceptions for high-achieving players, and this completely changed how teams build rosters: instead of buying one star, many teams allocate budget across three above-average positions. A financial report that does not state the salary cap mechanism and its effective date cannot explain any transaction.

This is also where I hold a fairly firm professional view: transfer valuation models tend to overrate young potential and underrate dressing-room chemistry. Chemistry is a variable that cannot be measured in a spreadsheet, but it affects results directly, and it only becomes visible after the deal is done.

The sixth layer is rules and governance. Transfers, registration, contracts, protection of minor players, and competitive integrity. South Korea is a market with a long memory of match-fixing in esports, and that memory shapes how leagues operate their monitoring mechanisms. For a data journalist, this is the layer that most needs concrete citation: the effective date of a rule, the number of cases already adjudicated, and the scope of regulation. Without those three, the rules section is just a re-translation of a press release.

The seventh layer is the risk profile. Injury, burnout from schedule density, personnel risk when contracts expire simultaneously, and systemic risk when a region depends on a handful of top teams. Risk only means something when assigned a probability and an impact level. A line reading “high risk” without a probability is an exclamation, not an assessment.

The eighth layer is public narrative and expectation. Market expectation is usually built from a few moments replayed many times, and it deviates from reality along a fairly stable pattern: after a big win, expectation rises faster than roster quality; after a loss, expectation falls faster than the real decline. The gap between those two curves is where an analyst can create value, because it is a gap that can be quantified by comparing actual results against expectations built in advance.

The ninth layer is industry transmission: from publisher, through streaming platforms, to sponsorship, hardware markets, and grey zones. This is the longest and slowest layer. A change at the publisher level takes months to reach sponsorship, and several seasons to reach academies. It is therefore almost never where analysis begins, but where it ends.

The three most neglected variables when all nine layers are combined are these: the stable lag in pick-ban data after each update, the gap between game win rate and series win rate, and the effect of losing a live crowd on home teams. All three are measurable. All three rarely appear in a report with handsome tables.

Reading back through those nine layers, one trap stands out above the rest: the more polished the template, the harder the gap is to see. Forty tables create a sense of completeness. A reader skims, sees headers, sections, and tables, and assumes work sits behind them. But formal completeness is the cheapest thing to produce, and the easiest thing to disguise an empty input with.

There is one methodological error I have made before and still see every season: turning correlation into causation. In esports it usually appears as “Team X won because they controlled objectives”. But objective control is often a consequence of being ahead, not a cause of winning. Reversing causality in a chain like that produces a conclusion that sounds very reasonable and is entirely wrong. The only way out is to build the evidence chain chronologically at game level and check which variable appeared first.

The second error is turning missing data into a conclusion. When a cell returns empty, the writer's natural reflex is to fill it with language: “no precedent exists”, “insufficient basis”, “needs further monitoring”. Those three phrases sound cautious but are in fact claims about the world, and those claims have nothing behind them. A properly recorded empty cell should read as no data available, with a reason: no source, source not public, or source not credible.

The third error, and the most dangerous for a specialist writer, is letting a risk forecast slide into a verdict. I once wrote about a footballer's decline after injury based on positional data, and the prediction was right. But precisely because it was right once, the writer starts believing certainty is available. A model only produces a probability distribution. The correct sentence is “if running distance keeps falling at this rate, performance is likely to decline over six to eight weeks”, not “this player is finished”. The collapse of a giant always begins with a fragile xG, and a data writer should describe the fragility, not announce the date of the fall.

In esports these three errors resonate more strongly than in football, because the pace of updates outruns the pace of data publication. An update can hit the competitive server while the previous stage's statistics are still being compiled. That means a window always exists in which nobody holds complete data, and inside that window, the nine-layer template becomes a tool for looking busy.

The other way to use a template is as a map of gaps, not a map of conclusions. Each layer gets marked in one of three states: data exists and is verifiable, data exists but is unverified, and no data exists. Only the first state is allowed to produce conclusions. The other two produce a to-do list.

Done properly, an empty report stops being a failure. It becomes an honest inventory of what this industry publishes and what it withholds. And that is often the most valuable thing a data journalist can hand a reader.

Next cycle, I will put three questions to anyone sending me an esports analysis: where is the raw data, what is the exact calendar date, and which of these nine layers is allowed to produce a conclusion. If there is no answer to the third question, the rest is decoration.