Trends & clusters
Every game is decomposed onto four axes read from its title and description — subject (what it is about), mechanic (what you do), format (how it is packaged), and meme (the viral trend or franchise it rides on) — then grouped by tag per axis. These clusters, not any genre label, are the unit a forming trend is measured on. Alongside size, each cluster carries its breadth: the effective number of games its live players are spread across. A cluster with high CCU but breadth near 1 is one hit wearing a genre's name, not a trend.
Where's the opening
Each cluster read as a market: its audience (CCU) against how that audience is spread (breadth). A big audience owned by one game is a moat; the same audience shared across many games is a contestable lane. The verdict is derived from the numbers, not hand-written. Sort by any column.
Subject
Mechanic
Format
Meme
The viral trend or franchise a game rides on, pulled from its title and description. Unlike the other axes this one is partial: most games reference no meme and appear in no cluster here, so counts are a subset of the pull, not the whole of it.
A cluster is a canonicalizer bucket, not yet a semantic one. Clusters group on
a lightweight singularization of the tag string, so near-synonyms that are not
spelling variants stay separate — soccer and football are two clusters, as are
role-playing and roleplaying. Read a cluster as a trend allowing that a real
subject can be split across a couple of labels until semantic clustering replaces
string canonicalization.