Shipping Phaser at scale.
Running a Phaser 3 platform with 2,500 daily players and 50+ titles in production teaches you what does and doesn't scale. A few notes from the trenches.
Asset budgets matter more than engine choice.
The single biggest performance lever is total bytes loaded per session, not Phaser version, not WebGL settings. We hold ourselves to a hard ceiling per title and tune sprite sheets, audio compression, and font loading against that ceiling. Every title that breaks the ceiling gets a fresh pass before we ship.
Live ops is a different sport from launch.
Launching a game and operating one are different crafts. Launch is about novelty, polish, and the first thirty seconds. Live ops is about week-30 retention, seasonal cadence, and giving the world something new to react to. Different metrics, different reviews, different design conversations.
State must survive everything.
Players close laptops, lose connection, and switch devices. The platform's job is to make that invisible. Player state lives server-side, gameplay state is recoverable on reconnect, and save-anywhere is the default. Every regression here is felt for weeks.
Telemetry is a feature, not a side effect.
We instrument before we tune. Crash signals, drop-off funnels, economy faucets and sinks: all of it. The instrumentation has caught more bugs than QA has.
The engine gets you to launch. Operations get you to month three.