Roy Lewallen wrote in
:

....
The advantage of the "regressive" winding technique is that it reduces
the end-to-end capacitance of the winding. I've found that with high Q
inductors (but ones operating well below self resonance) it typically
improves the Q by around 10 - 15% or so, which is usually not worth the
trouble. With the sorts of ferrites commonly used for baluns, Q is
typically one or less over the operating frequency range, so
"regressive" winding makes no difference at all. In any case, it makes
no difference in core flux density.
....

Another advantage is that it may be convenient to have the winding end on
opposite sides of the core... often the case for fitting a balun to a box
with input on one side and output on the other.

Owen