Found it: the puzzle generator had a "parity check" that only allowed clicking each tile once. On a 3x3 grid (9 tiles), this capped all puzzles at 9 moves maximum, even though levels 10-20 were supposed to require 10-20 moves.

The root cause: We were using binary logic (clicked/not-clicked) in a game with 3-color cycling. With 3 colors, each tile can be usefully clicked up to 2 times before returning to its original state.

fix v1.87.0: Changed from binary parity to tracking actual click counts, allowing each tile to be clicked (colors - 1) times. This enables up to 18 moves on a 3x3 grid, exactly as pyt correctly calculated!

Now levels properly progress: - Levels 1-18: 3x3 grid (max 18 moves) - Levels 19-32: 4x4 grid (max 32 moves) - Levels 33+: 5x5 grid (max 50 moves)

Thanks for playing!

As you say, Gaussian elimination can be used for a more systematic approach.

It reminds me of the dual n-back game, where you had to remember N steps before of 2 things. You start with 1-back and progress once you "get it".

As it is now, I just couldn't "get" what was the required reasoning behind this puzzle, before I got frustrated and left. And the hints didn't give me nothing personally, because once you get it, it basically solves it, without actually helping you understand the reasoning process.

Also similar to Sudoku... if you start with a difficult one, you just get lost. You have to learn the reasoning tricks.

  - Levels 1-18: 3x3 grid (max 18 moves)
  - Levels 19-32: 4x4 grid (max 32 moves)
  - Levels 33+: 5x5 grid (max 50 moves)