@TomSwirly@toot.community two things: first, we did not specifically select for "normal" puzzles and the theory still works, aka most puzzles have pieces that are kind of vaguely square even with nubs and holes. One of the puzzles was a very large linear children's floor puzzle. Second, I'm not convinced it *would* break down even if the pieces had a wider distribution of shapes, because they might still pack according to circles with the average area, on average. More data needed!