They're not quite the same. If you swapped two vertices, you'd have to remove 4 edges. Specifically, for each of the two chosen vertices, you'd have to remove both of the edges which were connected to that vertex. Then 4 edges are added. On the other hand, in 2-opt only two edges are removed (and 2 new edges are added). Still, swapping two vertices does also seem like a good method of iterative improvement. I couldn't find any information about it online though, so my guess is that 2-opt is just better.