The first independent review of Salvage Swarm gave Level design a 2 out of 5. Not a content problem. A shape problem.
Salvage Swarm is a robot-salvage puzzle game. Crates arrive on conveyor lanes. Tap one to dock it, and its robot crew breaks apart every block of that color in a pixel picture. Robots only reach the lowest block in each column, so the order matters. Five bays. Don’t jam them.
The shape problem was that the difficulty curve wasn’t doing anything. Easy and hard levels sat side by side with no visible difference in how hard they actually were. So the fix was measurement: a simulated casual player that taps randomly within the rules, run against every level. From there the goal was simple to write down. The win rate should fall from one order to the next and never rise.
The fixes came in stages. Chains, pairs of crates that dock together and need two free bays. Mystery lanes, where only the front crate shows its color. Short-staffed orders, four bays instead of five. By the time the campaign hit 12 levels, the casual tapper was winning 25% of the time on the hardest.
Then the campaign expanded to 100 orders across 10 chapters, and the arc had to hold over a longer run. Easy orders stay at 82% or above. Hard orders run 45-72%. Super hard orders hit 10-38%. The finale, the last Super hard in the campaign, is pinned at 3%.
The daily puzzle has a floor of 10%. A daily should always be beatable by someone who shows up that day. The finale is a different situation. It sits at the end of 100 orders, with every mechanic already introduced and every hard order already solved. Three percent is what the difficulty measurement says the finale should be, not a ceiling someone set before they knew better.
The first 10 levels of the game start out easy. Almost too easy. This is to get the player used to the game mechanics. It gets progressively harder each chapter. That’s not an opinion. It’s a measured fact.
The finale at 3% isn’t a problem to fix. It’s the design.