After cracking my brain for 2 weeks in the search of the polynominal solution, I decided to use my observations to code the most optimized random algorithm I could think of.

Surprisingly, I didn't even have to use simulated annealing: hill climb was enough. I also have 2 optimizations that I planned to implement but didn't have enough time.

I was expecting very strong tests considering all the dramatic comments from coders way better than me.

Maybe it's only a matter of time before admins add new tests that crack my solution, but I can't think of a single test that is strong enough to survive all optimizations...

I'm extremely curious about the polynominal solution, so maybe I will continue thinking about this problem... But it's so incredibly hard...