Parking my idle game: when the maths of idle growth beat me (for now)
I'm putting Genesis on the shelf. Not deleting it, not giving up on it, just parking it until I know how to finish it. This post is about why, because the reasons are more useful than the game is right now.
The loop that ate itself
The clearest sign something was wrong came from a long test run. After a day and two hours of play and 341 Extinctions:
- the median time between resets was 3 seconds,
- real gameplay had effectively ended around Extinction 40,
- the player was sitting on 465,793 DNA with nothing worth buying: the next sink cost a million.
A reset that takes three seconds isn't a decision, it's a reflex. The prestige loop, the heart of any idle game, had collapsed.
What caused it wasn't one bad number. It was four separate carry-over decisions, each reasonable on its own: keep a bit of this, carry a bonus from that, let this upgrade survive the reset. Stacked together, they meant a reset removed no consequence at all. I wrote it down in my notes as one design error made four times. No amount of tuning could reach it, because it wasn't a tuning problem.
That led to a restructure into chapters, where nothing crosses a chapter boundary except a fixed amount of starting currency. It fixed the collapse. It also reset every target I'd been tuning towards.
Too many tools, measuring the wrong player
By the time I stopped, the project had 87 probe scripts and 79 npm commands, each written to answer one question at one point in time. I deleted 61 of the scripts and 46 of the commands, and wrote down the question each one answered so I could rebuild it rather than reinvent it.
I also deleted the whole-arc tuner, because the honest summary of its history was this: every bake had measured a player who couldn't exist. A bot buying from tabs it couldn't see, a bot that never slept, an objective the tuner could satisfy by deleting the game.
No finish line
The real problem is the one I flagged in the design post: the game has no ending. Idle games traditionally don't. But "no ending" meant there was never a point where the game was done, only another chapter to tune and another curve to adjust. I'd been "nearly ready to release" for weeks.
What I'm doing instead
Two things.
First, I'm going to make one game properly: polished, finished, with a real ending to its design, and use it to answer a question I've been avoiding: does paid advertising work for a small game? That's Sovereign, and it's next.
Second, Genesis waits. When I come back to it, I want three things decided before I touch a number:
- It ends. A finale, a finish line, something to tune towards.
- Simulated players play like humans, checking in a few times a day, not a bot that never sleeps.
- The scope is fixed. Two chapters, done, rather than three, nearly done.
I learned a huge amount from this game: closed-form maths, soft caps, simulation, auto-tuning. I'll be back for it.
