Reading an esports event through nine layers of data: from patch notes to cash flow

In a semifinal at an international event this season, the winning side dominated in kills — 42 to 28 — yet its gold difference per minute was negative across the first two games. The stands roared, the scoreboard lit up, and I stayed behind with the data file, asking myself: if kills did not build the win, then what did? The answer is not a single number. It lives in nine layers of information that viewers usually skip because of the noise of the moment. I wrote my blog from a rented room in Nha Trang; now probability takes me everywhere. In the early days I also thought esports analysis was just counting kills and watching highlights. But after many seasons of tracking, I learned one thing: the match ends, but the data stays. And it is precisely the data left after the applause where the truth of a match resides. The prerequisite of any esports analysis is identifying the correct game title. You cannot apply the same ruler to League of Legends, DOTA 2, CS2, Valorant or Honor of Kings. Each title has its own tempo, resource system and scoring logic. A model built for DOTA 2 is meaningless if aimed at Valorant, where round economy and kills decide in a completely different way. Skip this step, and every conclusion afterward is mere inference. The first layer is patch and meta. I do not read an update by the feeling of "this patch buffed that champion", but by three columns of numbers: win rate, pick-ban rate, and average match length after the patch hits the tournament server. A small stat change can push a champion's pick rate from 12% to 60% in just two weeks. Conversely, some patches look huge but barely shift team composition, because they only touch a rarely used group. In a regional event last season, I noted a top-lane champion with a 54% win rate but only an 8% ban rate. That was a sign of an underexploited meta, and exactly two weeks later its ban rate jumped to 47%. Telling these two kinds of change apart is the line between a headline reader and a numbers reader. The second layer is tournament format. BO3 differs from BO5, upper-lower bracket differs from a points-based group stage, Swiss differs from single elimination. I once proved with data that under single-elimination formats, the favorite's win probability drops sharply compared to a BO5 series. Not because the stronger team got weaker, but because fewer games let variance overpower strength. A lower-ranked team only needs two correct games to advance. Format is not neutral; it is a variable with its own weight. The third layer is teams and players. Here I separate three things: paper strength, role fit and roster chemistry. A marquee signing can raise mechanical stats while breaking the team's communication rhythm. In esports, shot-calling and the rapport between a jungler and a mid laner matter no less than individual skill. I have seen a team spend the most in a transfer window yet slide in results, simply because two stars refused to yield the shot-calling role to each other. Data cannot measure ego, but it measures the consequence: successful coordination plays drop, reaction time in teamfights rises. A young player may post a high creep score per minute but also a high solo-death rate; the second number is what a coach needs to look at. The fourth layer is the regional picture. China, Korea, Europe, North America, Southeast Asia — each region has its own style and its own strength per title. Some regions dominate DOTA 2 yet lag in Valorant. Tracking the flow of imported players and exported talent gives me early signals of a power shift. When a region starts importing coaches en masse, that is usually a sign it is falling behind tactically and needs to buy experience. The fifth layer is club finance. I look at four lines: sponsorship revenue, league distributions, salary expenses and owner capital. A team can be on a winning streak yet still be unbalanced if its payroll far exceeds its income. In this industry, sponsorship and rights fees are two legs, while payroll is a mouth that never stops opening. When an owner pulls out, an entire roster can dissolve within a season. I once tracked an organization whose payroll was three times its sponsorship revenue; two seasons later, it disbanded its main roster. I treat "mandatory purchase" loan deals as a worrying signal, because they tie the hands of small teams and turn them into farms of semi-finished products for the giants. The sixth layer is rules and governance. Competitive integrity, transfer rules, contracts and even disputes around publishers all sit here. A late sanction, a rule change mid-season can upend the entire calculation. I always read the rulebook carefully before analyzing, because the rule is the least-mentioned variable with the greatest weight. The seventh layer is the risk profile. I split it into competitive, financial, personnel, rules, public-opinion and systemic risk. Each has its own probability and impact. A wrist injury to a star player, an off-stage scandal, or a declining game lifecycle are all risks that can be partly quantified. The eighth layer is the media narrative. Labels like "new king", "dynasty", "all-domestic roster", "last dance" pump heat into an event, but they do not always rest on a data foundation. I separate heat from strength by comparing social-media reach against actual form metrics. When a team's reach doubles while its form metrics stay flat, I start raising questions. The ninth layer is industry transmission. From publishers upstream, through clubs and streaming platforms midstream, to sponsorship and derivative markets downstream, every change spreads along the chain. A scheduling decision can shift the value of an entire ecosystem. What I want to say against the crowd is not about those nine layers, but about how people use them. People call me a "number freak"; I take that as a compliment. But I also have to remind myself: correlation is not causation. A team winning more after changing coaches does not mean the change produced the wins. Maybe the schedule got lighter, maybe the new patch suits them. And above all, the biggest risk in analysis is analyzing empty data. If there is no game title, no team, no single information point, then every conclusion is fabrication. A good process must have a minimum gate: at least one title, one entity and one data point, before any judgment is allowed to be born. An empty arena does not need spectators; it needs an analyst willing to look. As the season enters its final stretch, I will keep betting on the signals the crowd overlooks: the form curve of a young player, an unfamous patch, or money quietly withdrawing from an organization. The match ends, but the data stays. And the question I carry into the next round is simple: if you could keep only one layer of data to read a season, which one would you choose?