Recognizing Patterns · Intelligence Suite

The Community Mobilization TrapCommunities mobilize fast, but rarely convert into something durable.

Four live-service games acquired forty-eight million players and shut down anyway. Two governments fell within a week of mass protest and lost control of the country within six months. Different domains, same failure: both mobilized people fast and cheap, using tools built for exactly that, and neither converted that mobilization into something that lasted.

This report tests one hypothesis against both: coordination has gotten radically cheaper, but the organizational work that turns a crowd into a durable institution — a retained playerbase, a functioning government — still takes years, still costs money, and cannot be skipped.

Part I tests it against four game shutdowns. Part II tests the community-signal literature the games industry uses to justify skipping it. Part III tests it against Nepal and Madagascar's 2025 uprisings.

Publication Recognizing Patterns
Published August 2026
Reading ~35 min total
48M
Players acquired by four titles that shut down
0.15x
Median Steam wishlist to week-one conversion
11 of 12
Failed cases that shipped a player-authoring layer
Year 3
The horizon nobody underwrites

The 48M total is not a unique-user count: the four titles used different definitions (registered players, lifetime players, and one three-week total) and are not additively comparable.

Reading order Part I establishes the acquisition-retention split. Parts II and III depend on it. Read in order.
I
The Acquisition Trap
Forty-eight million players, four shutdowns, and the participation-infrastructure hypothesis that dies on contact with the launch dates. Plus what actually separates the games that lasted.
Open Part I →
II
The Signal Is Not the Cause
Product produces sentiment; sentiment produces conversion. Wishlist dispersion, Valve's own position, the discourse-versus-audience gap, and a correction to a position taken in an earlier report.
Open Part II →
III
The Succession Problem
Nepal and Madagascar from primary sources, a century of campaign data, the literature that contradicts this report, and the year-three diagnostic that comes out of it.
Open Part III →
Quick Reference: Jump to a case
XDefiant & MultiVersus
Acquired 31M · Both shut down
Worlds Adrift
Editor shipped a year early · Peak 2,057
Helldivers 2
Launch peak · Largest playerbase in sample
The Wishlist
R² 0.49 · 10-20x dispersion
Nepal
Four days · Then eighteen months
Madagascar
Six days · Movement not consulted
Participation infrastructure is neither necessary nor sufficient for retention. Eleven of the twelve failed cases in this report shipped a genuine player-authoring layer, and Worlds Adrift shipped its island editor more than a year before the game itself, peaking at 2,057 concurrent. Meanwhile three of the longest-lived games in the sample, Valheim, Path of Exile and Helldivers 2, never shipped one at all. What separates them is sustained content cadence over years, which is unglamorous, expensive, and almost never funded at greenlight.
This report was commissioned to test a thesis and a team was tasked with destroying it. Six of eight component claims fell. The replacement claim then failed its own second review for selecting on the outcome, and a third pass falsified the replacement's central mechanism. All three corrections are in the text rather than in a footnote, because the corrections are more useful than the original argument was.

The instrument that would settle most of this argument is the one nobody publishes: what a launch cost to market, next to what it earned. Until that number exists, every thesis in this space is inference dressed as measurement, including this one.

Abbas Saleem is a Principal Consultant at Llama & Griffin, advising game studios, publishers, and investors across MENA, South Asia, Southeast Asia, and Europe. He writes Recognizing Patterns: gaming industry intelligence 12 to 24 months before it becomes consensus. LinkedIn