[This explanation _begs_ for some pretty animations.] Consider a _single-level_ rope made of 2 individual continuous/unbroken fiber strands. We assume that these individual fiber strands are completely limp, with no resistance to bending, _no resistance to twisting_, and a typical linear spring constant stiffness to longitudinal stretching. If both untwisted fibers are of length L>>1 and radius r, then twisting them to achieve a helical angle of alpha, where alpha=0 means zero twist, will shorten the overall rope, and the twisting torque can be calculated by how much the rope length will change with a given change in twisting angle. Let us consider a unit disk in the XY plane, and a parallel unit disk around (0,0,L). We attach fiber #1 at (1,0,0) and at (1,0,L) and fiber #2 at (-1,0,0) and at (-1,0,L). We assume r<1, and both fibers are straight and parallel. If we then rotate the disk at Z=L by a certain angle alpha, then not only will the entire rope be twisted by alpha, but each fiber strand will also be twisted by alpha, but since we have assumed that torquing an individual fiber encounters no resistance and does not change its length, the overall behavior of a one-level rope of 2 strands is to untwist (unravel) under load. Let us now consider a _two-level_ rope, twisted from 2 strands, each of which is itself twisted from 2 strands of individual fibers. Let us start with all 4 fibers straight. Fiber #1 goes from (1,r,0) to (1,r,L); Fiber #2 goes from (1,-r,0) to (1,-r,L); Fiber #3 goes from (-1,r,0) to (-1,r,L); Fiber #4 goes from (-1,-r,0) to (-1,-r,L). If we now rotate the disk at Z=L by angle alpha, then all 4 fibers will wrap around each other in a _single-level_ rope with 4 strands, and since all the fibers are wrapped in the same direction, untwisting it by angle alpha will lengthen the rope, and thereby impart a torque tending to unravel the rope. But consider now what happens if we simultaneously twist the 2 lower (smaller) level strands by +alpha and twist the 2 higher (larger) level strands by -beta, such that the untwisting torque of -alpha is matched by the twisting torque of +beta. The overall tendency of the entire 2-level rope is to change length (with some sort of spring constant k) under load, but to have _no net twisting torque_. (We note that fixing the ends of the rope so that the inner strands cannot rotate against one another is essential, else the inner rotation will allow the outer torque to become non-zero, and the rope will then unravel itself catastrophically.) If we have a 2-level rope made up of m inner strands and n outer strands, and the inner strands are round fibers of radius r, then we can calculate the overall properties of the rope. (A slightly more accurate model would have the radius r of each fiber change slightly under longitudinal load, so that the _volume_ of a particular piece of fiber remains constant when it is stretched.) We can also relax the constraint on continuous/unbroken fibers. Since loading a 2-level rope increases the fiber-to-fiber friction (as in our previous untwisted WDSmith model), under normal operating conditions this friction is enough to keep the individual fibers from slipping against one another, so that a break in one individual fiber will only reduce the carrying capacity of the rope by a factor of 1/(m*n). There are still some interesting questions about what different "tiling patterns" occur with different integers m & n, where "tiling patterns" means how the various strands at level 0 interact with the strands at level 1 where they meet (and possibly nest in indentations), with depending on the different angles of twisting alpha & beta.