A and B both enter the same face of the core, therefore they are both starts. Or both ends, but just pick one direction as convention for argument's sake.
Another way to think of it: slide the turns around, until they overlap. You can reverse the position of B and D by loosening the winding, and pushing the turns over and under each other,
without ever passing an end of the winding through the core. This does not invert the phase of B and D, B is still the start and D is still the end. Now, while the turns are still loose, rotate them around the core until on top of A-C. You will find that B and A line up perfectly, and the two windings follow identical paths (once you snug up the wires), all the way to C/D. You can merge them together mentally as a single cable (this is called bifilar construction, when done intentionally), or connect them together and see that the inductance is the same as the single winding alone.
The key factor is, how many loops are made around/through the core, and in which direction, front to back or back to front. (A starts at the front, and C ends at the back.)
Incidentally, a "necklace" (a single toroid core on an otherwise seemingly straight wire) is one turn. How? Well, sooner or later, if we are to measure a current through that wire, it has to complete a loop. It doesn't matter that that loop is the size of the universe, or wrapped tightly around the core -- it still makes one complete turn.
Ultimately this connects with topology and winding number (knot theory).

Tim