ourplayhub ourplayhub

Launch Report: STARCLAIM on CrazyGames — and the Level 6 Wall

STARCLAIM Devlog · July 22, 2026

STARCLAIM spent the last few weeks on CrazyGames, one of the biggest web-game portals in the world, as part of its launch evaluation program. For a tiny indie hub, that's the moment a game stops being played by friends and starts being played by strangers — hundreds of them, with zero patience and no reason to be kind. The run is over now: portals grade you on retention, the bar is brutal, and we didn't clear it this time. But the data those weeks produced permanently changed the game. This is the honest launch report: what ~300 players did, where they quit, and the balance patch their data forced us to ship.

Launching on a portal is a firehose

On our own site, traffic arrived one curious visitor at a time. On CrazyGames, the launch queue put STARCLAIM in front of hundreds of players within days. Portal players behave differently: they click, they judge in seconds, and they leave without a word. The only way to learn anything is telemetry — anonymous play events that tell us which campaign level a match started on and how it ended. No accounts, no personal data, just "someone started level 4 and won."

Before we could trust that data we had to fix it. Portal sessions are hostile to analytics: ad breaks, tab switches and flaky mobile networks were silently swallowing events, which made the funnel lie to us. We rewrote the logging as a write-ahead queue — every event is saved locally first, sent after, and retried until the server confirms it. Boring plumbing, but every number below exists because of it.

Reading the funnel: where players actually quit

Of the players who tried the new Campaign mode, 68 started level 1. Then, level by level: 43 reached level 2, 36 reached level 3, 27, 23… a steady, expected drain — casual players sampling a game and moving on. But one number refused to fit the pattern: the loss rate. On levels 1–5, players lost between 0% and 9% of their matches. On level 6, the loss rate spiked to 23% — and then dropped back to nearly zero on levels 7–10. That's the signature of a wall: it wasn't that level 7 was easy, it's that level 6 had already filtered out everyone who couldn't beat it.

The culprit was one number

Campaign levels are generated from a single pure function — give it a level number, it returns the map, the AI temperament and the starting forces. Digging into that curve, the problem was embarrassingly concrete: the player's starting units took a hard step down (60 → 50) at level 4, exactly while the AI's starting units were climbing. By level 6 the player-to-AI power ratio had quietly slid to about 1.6 — the tightest squeeze in the entire early game, hitting players who had only been commanding fleets for ten minutes.

The fix is what we should have shipped the first time: one smooth, monotonic slope across levels 1–10. The player's starting fleet now glides down (60, 60, 58, 56, 54, 52, 50…) instead of stepping, and the AI ramp was gentled to match, lifting the level-6 ratio from 1.61 to 1.79. Late-game levels are untouched — the climb past level 10 is supposed to bite. Our balance simulator (thousands of automated matches against a deliberately weak bot) now gates every level from 1 to 10, so a hidden cliff like this can't sneak back in.

Did it work — and what's next

If you've read our campaign devlog, this is the other half of the story: shipping the feature was the easy part — watching real players collide with it is where the actual game design happens.

Join the funnel: ▶ Play STARCLAIM — free, in your browser, no install. Level 6 is waiting, and it's fairer now. We'll be back on the portals with a stickier game.