So the point is that always only exactly 2 elements are compared and so you first have to split everything into groups of 2. Seems very inefficient for larger datasets, since you need to handle every single item over and over again and compare so so often. But not a sorting and comparison expert, so no idea if human sorting logic applies at all.