When Cosplay Wears the Esports Label: The Classification Gap Eroding Gaming Data

One mid-regular-season morning, while opening an esports content digest on a Vietnamese portal to cross-check figures for a transfer-valuation report, I found an anomaly sitting among regional qualifier bulletins and transfer lists. It was a cosplay photo set of the character Shimakaze from the mobile game Azur Lane, accompanied by praise for her "easily recognizable design" and "ability to transform through many outfits." No tournament. No team. No player. No patch. Just a photo set. And it sat neatly inside the esports category. What made me stop was not the photo set itself, but its position in the data structure. When cosplay content is automatically filed alongside transfer news and qualifier results, the error is not in the photo set. The error is in the label attached to it. For a data person, a wrong label is more dangerous than a bad number, because a bad number only corrupts one conclusion, while a wrong label corrupts an entire way of seeing. Azur Lane is a character-collection gacha game developed by Manjuu and Yongshi, operated on a cycle of new banners and outfits, not on a cycle of balance patches like competitive titles. Shimakaze is a destroyer of the Sakura Empire faction, notable for her rabbit-ear design, sailor uniform, and "warrior but cute" tone. Those traits serve commercial and image-propagation purposes, not competitive ones. Put differently, what governs this IP's rhythm is the banner and skin rotation, not the balance patch. The content I read is a product-introduction piece: a cosplay photo set by a content creator, presented by a portal with a supportive stance and a promotional purpose. Within its structure, no entity belonging to the competitive ecosystem exists: no tournament, no team, no professional player, no coach, no contract, no transfer, no club finance, no governance issue. Those categories are not missing data — they do not exist structurally. This is the most important distinction between "insufficient information" and "inapplicable." When a cosplay article enters an esports analytics pipeline, it does not merely dilute content. It corrupts data. If a digest claims that esports content volume rose by some percentage that month, but hundreds of cosplay sets were tagged via keyword adjacency, that growth figure does not reflect the competitive industry's development. It reflects the inflation of an empty category. And as I keep telling colleagues in Berlin: numbers never lie — only the reader's heart turns them into lies. The gacha IP flywheel To understand why a cosplay set can slip into an esports feed, one must understand how a gacha IP operates. Unlike a competitive game, where value lies in skill and match results, a gacha IP's value lies in the emotional pull of its characters. Shimakaze was not designed to be strong in a competitive meta — she was designed to be recognizable, easy to redraw, easy to cosplay, and easy to become the center of fan-produced content. That design is a business decision, not a balance decision. This is a three-tier flywheel. The upstream tier is the publisher, creating characters and outfits. The midstream tier is the content-creator community — cosplayers, artists, video editors — turning characters into viral cultural products. The downstream tier is the fandom, where emotional attachment converts into revenue from skins, merchandise, and events. This flywheel feeds itself, and its rotation speed depends on no tournament whatsoever. Looking at that flywheel, I see something the esports industry often overlooks. An IP can sustain vitality for years without an official competitive system. It only needs a character recognizable enough for the community to keep recreating. From a transfer-market perspective, this is a lesson about value flow: value does not only flow through the competitive channel; it also flows through the fan-content channel. And these two flows have entirely different decay rates. The economics of fan-content placement The cosplay set did not appear on the portal by chance. Behind it lies a clear economic logic. Gaming and esports portals compete fiercely for traffic. Competitive news has a loyal but narrow audience that fluctuates with the tournament calendar. Cosplay content has a broader, steadier audience and spreads more easily on social media thanks to its visual nature. Hence a hybrid model emerges. The portal runs two content streams in parallel: a competitive stream to preserve editorial credibility, and a fan-content stream to preserve traffic. In media-engineering terms, this is a reasonable decision. In data-quality terms, it is a latent disaster, because both streams often get merged under a single label. When I asked an editor in Vietnam about categorization, the answer left me silent for a few seconds. Labels are usually applied automatically based on keywords and related articles. A cosplay article about a game character, sitting next to PUBG links, gets clustered with esports by the algorithm. No one deliberately mislabels. The algorithm simply sees the word "game" and sees esports links, then infers on its own. This is where two kinds of error must be distinguished. Intentional error is when a portal deliberately labels non-esports content as esports to lure traffic. Systemic error is when a classification algorithm misjudges. In this case, the evidence leans toward the latter. But the data consequences of both errors are identical, and the analyst downstream has no way to tell them apart without manual auditing. The real patch: the banner cycle In game analysis, I usually start by identifying which patch is in force. With Azur Lane, there is no balance patch to analyze. What governs this IP is the banner and skin rotation. This is a crucial point, because it changes the unit of measurement entirely. For a competitive game, the units are win rate, pick rate, ban rate, and per-patch performance metrics. For a gacha game, the units are character recognizability, volume of fan content produced, and community willingness to spend. These two unit systems cannot be converted into one another. Attempting such conversion is a fundamental methodological error. This explains why meta analysis is inapplicable to this content. No meta direction, no beneficiaries, no losers, no win-rate data. Not because the source lacks information, but because in this genre no competitive meta exists. Once again, this is the difference between missing data and inapplicable analysis. The decay coefficient and the lifecycle of cosplay content I once built a tool called the Decay Coefficient to measure how a team's form declines over time. Its principle is simple: everything decays, differing only in speed. I applied the same principle to media content, and the results were revealing. A transfer rumor has a short half-life — it flares for a few hours then fades. A tactical analysis has a longer half-life, possibly surviving weeks if it contains original insight. A cosplay set has a medium half-life — it spreads strongly in the first few days, then depends on whether the IP has new events to resurrect it. These three content types have three different decay curves, and merging them into one category makes any lifecycle analysis impossible. The source's analytics tool provides no engagement metrics. No views, no engagement rate, no shares. The praise "a rather impressive transformation" is a qualitative claim with no unit of measure. In my language, that is an unlabeled data point. And every crisis is unlabeled data — including small crises like a photo set placed in the wrong slot. The decay coefficient of cosplay content depends on the IP's event calendar. If Azur Lane releases a new Shimakaze skin or holds an anniversary, the old set can be resurrected and re-spread. If not, it sinks into oblivion within a month. This is why I suspect the posting date may coincide with an IP event window, though the source gives no clear timing signal. That is a hypothesis to verify, not a conclusion. The real signal lies at the edge: the Himass case While the body text covers only cosplay, the related-links block contains a signal of an entirely different nature. It is the PUBG Asia Stars event, featuring a Vietnamese player named Himass facing a possible suspension, along with developments around KRAFTON's response and a dispute among the parties involved. This is genuine esports content. A player, a tournament, a publisher, a disciplinary risk, a governance dispute between two countries. Every category the body text lacks — tournament, player, governance, public opinion — appears here. The contrast between the two sections of the same website is a clear marker of the source's hybrid nature. That a real esports signal sits embedded in the related-links block of a cosplay article says much about editorial strategy. The portal operates as a hybrid model: using light content to attract mass traffic while maintaining genuine competitive news to retain in-depth readers. The problem is that when both streams flow into the same data pipeline, the downstream analyst can no longer distinguish signal from noise. If the goal is to analyze the Himass case, I must separate it and treat it as an independent matter. A cosplay article cannot serve as a foundation for tournament analysis. Mixing the two into one analytical package is a methodological error, not a data limitation. The portal's multi-title strategy Looking at the link structure, I see the outline of a familiar strategy. Southeast Asian portals typically operate across multiple titles: one day covering League of Legends, the next PUBG, the next a new mobile game. The goal is maximum coverage and retaining as many reader groups as possible. This strategy has upsides and downsides. The upside is resilience when a title cools — if League of Legends declines, other titles can carry traffic. The downside is blurred identity. When a portal publishes content of all kinds, readers struggle to picture exactly what they are reading. And the output data loses its purity. In the data field, I always repeat one principle: output quality cannot exceed input quality. If the input is a mix of content not rigorously labeled, the output can only be a beautiful but meaningless table. A cosplay set sitting in the same basket as qualifier news is a symptom of an input problem, not an analysis problem. I once saw an automated classification system in Europe merge Valorant tournament news with news about a game-adaptation film, simply because both contained the same keyword. The result was that a quarterly report from an investment fund completely miscalculated the regional esports market size. A wrong label is not a minor technical detail. It can lead to investment decisions that are off by millions. Structural comparison: competitive content vs fan content To see the mismatch clearly, I run a quick comparison in my head. A standard competitive analysis needs: a verifiable metric, a tactical context, a chain of evidence, a counterintuitive angle, and a forward-looking conclusion. A cosplay introduction needs only: a photo set, a compliment, and a reason for readers to click. The gap between these two content types is not just a gap in topic. It is a gap in method. Competitive content is built on a hypothesis - verification - conclusion structure. Promotional content is built on an evoke emotion - build goodwill - prompt action structure. Merging the two into one category merges two incompatible method systems. The consequence is that when an analyst reads an esports content digest, that person does not know what percentage is real analysis and what percentage is promotion. This ratio, if unchecked, will distort any industry-trend conclusion. An industry can appear to be booming in content volume while actually only booming in promotional content volume. This is a form of information inflation, and it is dangerous because it looks like growth. I always remind myself that a beautiful dataset does not equal a correct dataset. Formal precision can mask substantive distortion. In this case, the presentation is perfectly fine, but the category's substance has been warped from the root. Labels and algorithms: the SEO problem There is another layer worth digging into. Content labels do not only serve analysis; they serve search algorithms. For portals, tagging correctly is part of SEO strategy. A cosplay article about a game character can be tagged esports to reach a larger reader group, because esports is a high-volume, stable keyword. From a purely commercial standpoint, this is an optimal decision. From a data-integrity standpoint, it is noise-generating behavior. I call this phenomenon label inflation — a label has value because it distinguishes, but when applied too broadly, it loses its distinguishing power. In the end, a label applied to everything is a label that says nothing. As search algorithms increasingly emphasize "information gain" — the new value content brings readers — mislabeling becomes an increasingly risky strategy. A cosplay set brings no esports information gain. It brings aesthetic gain. Those are two different kinds of value serving two different reader groups. Forcing them into one category harms both. Modern search algorithms have begun penalizing content whose labels do not match its body. This means the broad-labeling strategy, though short-term beneficial, may backfire long-term. A portal labeling cosplay as esports may boost immediate traffic but trades away long-term topical credibility. This is a trade-off I always advise clients to avoid. Inverse valuation: why a wrong label is dangerous In transfer valuation, I always start with the reverse question: why should we not buy this player. The same principle applies to data: why should we not trust this label. When I pose the reverse question before an esports content category contaminated by cosplay, the answer emerges clearly. Three harms reinforce one another. A wrong label corrupts measurement metrics: if the cosplay share of the esports category rises, indicators like "esports interest" get artificially inflated. A wrong label corrupts predictive ability: a model trained on contaminated data will forecast wrong trends. And a wrong label corrupts cross-region comparability: regions with a high share of promotional content will look livelier than reality. These three harms do not merely distort one number; they distort a whole way of seeing. And for a data person, distorting a way of seeing is the gravest offense. Bending numbers to prove a story already written in one's head is the white-collar fraud of data work. For a data monk, forging one's own scripture is the worst mistake of all. The irony is that in many cases, no one intends to bend the data. Distortion is born from laziness in classification, from traffic pressure, from the habit of broad labeling. Small, individually harmless causes accumulate into a serious systemic problem. This is why I believe data quality is a cultural issue, not merely a technical one. Regional view: the Vietnamese market There is a small but notable detail. The article's author is named Tuấn Hưng, and related headlines carry a Vietnamese context. This allows me to infer that the portal serves Vietnamese-speaking readers. That is an inference about a media market, not about a competitive region. The Vietnamese gaming and esports content market has its own traits. It is smaller than the English and Chinese markets but very dynamic and fast-growing in users. In such a market, traffic pressure is greater and content mixing stronger. Portals must race to produce enough content daily, and in that race, rigorous classification is usually the first sacrifice. The cosplay and fan-content economy is fragmented by language community: Japanese, Chinese, Western, and Southeast Asian. Cross-pollination between these communities occurs mainly through image-sharing platforms, forming a structure unlike that of regional leagues. In that structure, a cosplay set can spread from one community to another without any competitive event as a bridge. This makes me wonder about a paradox. Esports is built on organized competition, with league systems and strict rules. Fan content is built on free emotion, with no rules. These two worlds share the same audience but operate on completely different logics. Their merging into one media category is understandable, but their merging into one analytical category is a mistake. Public narrative and expectation The current public narrative revolves around the idea of a beloved character stepping into the real world through a high-quality cosplay. This is a narrative at the fan-content tier, not the competitive tier. It is supported by aesthetics and character recognition, not by competitive results or verifiable achievements. Measured against data standards, this narrative lacks quantitative grounding. No views, no shares, no engagement rate. The praise of an impressive transformation is a promotional claim with no unit of measure. In public-opinion analysis, I often ask: does social heat match substantive foundation. In this case, that test cannot be run for lack of sentiment data. The expectation gap appears in three places. On photo-set quality, expectation is impressive and faithful, but there is no independent measure. On community response, expectation is drawing players and the anime/cosplay community, which is plausible but unverified. On character popularity, expectation is hard to ignore, which is fairly consistent with Shimakaze's established fandom standing. Notably, the claim about community cross-pollination — Azur Lane players plus the broader anime and cosplay community — is the only generalizable insight, pointing to the fan-content cross-pollination gacha IPs rely on. It is a real phenomenon, but here asserted without evidence. A correct but unsupported claim is still an incomplete claim. Industry transmission: which channel value flows through If I map the value transmission for this content, it does not pass through the esports channel. It passes through the character-IP channel. The upstream tier is publishers Manjuu and Yongshi creating characters and outfits. The midstream tier is cosplayers and fan-content creators. The downstream tier is the fandom and derivative markets like merchandise and events. No tournament, no broadcast viewership, no sponsors, no players. So the standard esports value chain does not apply. This does not mean the content is worthless. It has value, but that value flows through a different channel. As a market operator, I see a need to distinguish the two channels clearly to avoid mispricing assets. A gacha IP flywheel can spin sustainably without a competitive system. This is a truth the esports industry sometimes forgets. Not every game IP needs a tournament to survive. Some IPs live on the emotional pull of characters, nurtured by fan content rather than match results. At the macro level, the existence of such IPs raises a definitional question. If esports is defined as organized competitive video gaming, Azur Lane does not belong. But if esports is defined more broadly as any game-culture competition for attention, Azur Lane might sit at the edge. This boundary needs to be fixed, because all analysis depends on it. The cosplayer as a media node Another aspect deserves fair consideration. The cosplayer in the article is not a competitive actor, but nor is she meaningless. She plays the role of a media node in the fan-content network. Her value lies in the ability to convert character appeal into engaging content, not in competitive skill. In the content economy, these media nodes matter. They are connection points between publisher and community. They help sustain interest in an IP during periods without major events. And they generate a continuous content stream that keeps an IP from being forgotten. The problem only arises when their role is misunderstood. When a cosplayer is placed into an analytics table as if a professional player, every comparison becomes meaningless. You cannot measure a cosplayer by win rate, and you cannot measure a player by image virality. Each operates in its own unit system, and converting between them is methodologically impossible. I once saw a similar case in football. A club considered signing a player only because he had a large social-media following, despite poor match metrics. The club suffered losses in both performance and finance. The lesson is clear: media appeal and competitive ability are two different assets, and conflating them is an expensive mistake. Risk and classification auditing Globally, the risk of this cosplay article itself is very low. It threatens no competitive integrity, involves no club finance, has no personnel issue, no governance issue. The only content-tier risk is being seen as shallow traffic-bait, a mild reputational risk. But the systemic-tier risk is different. Mislabeling contaminates downstream datasets. One mislabeled article may be negligible, but thousands of mislabeled articles create a false trend. When that false trend is used to build models, the models produce wrong forecasts, and those wrong forecasts can lead to wrong investment decisions. I propose a periodic classification-audit process for portals. It involves randomly sampling articles tagged esports, manually checking whether they contain competitive entities, and calculating the contamination rate. If the contamination rate exceeds a threshold, the entire labeling system should be reviewed. This is tedious work, but it is the foundation of any trustworthy analysis. As a transfer-market operator, I am often asked why I spend time on such seemingly trivial matters. My answer is always the same: data is a chain, and one weak link can break the whole chain. Auditing labels is how we protect the first link. The counterintuitive angle At this point, I want to flip the problem once more. My initial hypothesis was that mislabeled esports tags are eroding industry data quality. But correlation does not imply causation, and a data person must defend against the lure of an attractive story. There is another possibility: the mislabeling itself is a symptom, not the cause. The boundary between esports and general game content has long been blurred. When a game like Azur Lane has a large player community, rich fan content, and community events, it also creates a form of broad competition for attention. The esports label then becomes an umbrella term sheltering everything game-related. If so, the problem lies not with a particular portal or article. The problem lies in the definition. When esports expands to cover cosplay content, it ceases to be a useful analytical category. It becomes a marketing label. And the irony is that I — who always demand label precision — must admit that the boundary I am defending may have been erased before this article was even written. The counterintuitive angle is this: the biggest risk is not a cosplay set labeled as esports. The biggest risk is that the industry has grown so accustomed to loose labeling that it no longer notices it is losing the ability to distinguish. When everything can be esports, nothing truly is. And numbers never lie — only the reader's heart turns them into lies. Forward-looking conclusion An empty-stadium summer, and I hear data dripping drop by drop. Those drops come not only from matches without spectators, but from unaudited content categories. In the next cycle, the signal to watch is not a particular cosplay set, but the shift in how portals define and label content. If the share of fan content in the esports category keeps rising without separation, every analytical model built on it becomes meaningless. The question for data people is not how much esports content there is, but what we are calling esports. And that question, regrettably, cannot be answered by a table.