One click, one chain

You get a single click, and you choose both where and when. Whatever the blast touches goes off in turn, and the level is cleared when enough balls are caught.

There are 100 levels. What rises is not only the number of balls but how large a share of the best possible click the level asks for.

Yes, and the target comes from a search rather than a guess. Before a level is kept, a grid of candidate clicks is simulated offline and the best result is recorded. The target is a fraction of that number, so at least one click is known to beat it.

The search is a grid, so it is a lower bound on the best click: there may be a better one between the grid points. That errs on the generous side, which is the correct side to err on.

Because the target of every level was measured in one language and has to hold in another. Positions, speeds and radii are integers, collisions compare squared distances with no square root, and the clock counts steps instead of milliseconds.

With decimals the two would agree almost always, and almost always is the worst possible outcome: the broken level would show up once in a few thousand rounds and nobody could reproduce it.

No. The animation advances in whole steps, and any leftover time is carried to the next frame. A 144 Hz screen draws the same round more smoothly, not a different round.

The furthest level you cleared is kept in your own browser. There is no account and nothing is sent anywhere.