JellySplit started as a complaint about Candy Crush. We loved the swap, but hated the randomness. So we kept the swap, borrowed a scoring rule from our last game Cell Division, and rebuilt the whole thing as a handcrafted puzzle.
The Candy Crush Problem
We played a lot of Candy Crush. Hours of it. But after a while, the same frustration kept surfacing — whether a level felt brilliant or brutal came down to the board the game dealt you. One cascade would pop twelve candies for free; an unlucky drop would end your run. The skill ceiling kept dissolving into luck.
We wanted a match-swap game where the puzzle was the puzzle — not a shuffle of tiles sitting on top of one.
Two Ingredients
From Candy Crush we kept the thing that hooked us: the swap. One jelly, one neighbor, one move. It’s the perfect verb for a puzzle game — cheap to understand, hard to master.
The scoring came from Cell Division, our previous game, where cells earn points based on how many of their neighbors share their color. More connections, more points. It’s a rule that rewards clusters, punishes isolation, and — critically — has zero randomness. Every jelly’s contribution is fully determined by the state of the board.
Drop the randomness. Keep the swap. Score the neighborhoods. That’s JellySplit.
Candy Crush × Rubik’s Cube
The best way we’ve found to describe the result: imagine if Candy Crush and a Rubik’s Cube had a baby. You swap like Candy Crush, but every puzzle has a fixed move budget and a target score — like the cube. No luck, no cascades, no “try again and hope.”
When you win, you won. When you lose, you know exactly what you should have done.
How We Build Puzzles
Every JellySplit level is handcrafted. We don’t generate random boards — we design them. The process looks like this:
- Start from the solution. We pick a final shape we want the player to reach — a tight green ring, a blue core with a yellow shell, a symmetrical split.
- Work backwards. We apply legal swaps in reverse until the board is far enough from the solution to feel earned, but close enough to feel fair.
- Set the budget. The move count is tuned so that sloppy play falls just short. Three stars means you found the intended line.
- Playtest until it hurts. A level only ships if at least one clean solution exists — and if the obvious first move is a trap. The best puzzles are the ones where the wrong swap looks irresistibly right.
Every puzzle has an answer. And every answer is a move you could have seen. We think that’s worth more than a thousand cascades.