I pushed the fix for my solution. As I suspected, the algorithm for identifying infinite regions was wrong, so in some situations my code included some infinite regions when looking for the maximum one. The new version now returns 3251 for your input, while the result for my input remains unchanged. I guess I was lucky here to get an input which didn’t trigger the bug in my code.

I struggled with this one more than I’m willing to admit, and ultimately made a brute-force solution which works reasonably fast and is not very ugly. You can find it here.

Today was a bit tricky… I scraped my elixir solution and wrote one in javascript (which would be trivial to translate to Elixir now when I know how).

Part 1.
I implemented a “flood fill” algorithm, relying on the fact that it is easy to enumerate all positions with distance N from a given position.

Part 2.
Step 1. Find the center of gravity
Step 2. ´Enumerate all positions with distance N from the center, count positions where the condition is true (sum all manhattan distances to the position). Continue to do this with distance N + 1, stop when the conditions fails for all positions in a level (N).

Looking at your code, it seems you assume that the finite region is a square, when it’s in fact a rectangle. So instead of finding one min-max pair, like you do in line 24, you need two pairs ((xmin, xmax) and (ymin, ymax)), and eliminate coordinates where x == xmin or xmax, or y == ymin or ymax.

I get the answer 3251 for your input.

Here is my solution.