What you're doing sounds fine. We did something similar, and what we got back was either 1) the obviously correct solution, 2) try-and-error soup, or 3) extremely complex over-engineered junk (we specifically told people not to do this, so double fail).

What most people object to is stuff that's just really time-consuming to do well. And/or stuff that gets rejected for silly reasons (typically requirements that weren't actually stated).

Or things like "please implement Conway's game of life in 30 minutes. START NOW".