Prack's Compendium 2 article explains a lot, which I paraphrase:
The antenna consists of paired vertical radiators made from wire. The system does not require
ground radials, nor does it require loading coils. It is fed directly with coax. It is an efficient
radiating system which presents a very low profile.
Dipoles are "electric" radiators, meaning they generate a very high E component in the near field. The E component is susceptible to ground losses. The H component, or the magnetic component on the other hand, is virtually lossless. Losses in the ground and in nearby objects are therefore minimized (as long as the objects are not ferrous).
The vertical ends of the antenna are 30 electrical degrees. [W0QL- I made my first one only 20 degrees and that is the "Four Foot Vertical". It works fine but down 1 s unit from the 30 degree antenna I made later. The 30 degree radiators are 6 feet tall for 20m.]. The configuration of the wire in the rectangular coil is such that the currents in the four vertical end wires are co-phased.
Prack continues to explain how to build one, including formulas.
My "Four Foot Vertical" is cut for 20 meters and the sides are 4 feet tall. The horizontal lengths are 32 feet and the bottom wires are two feet above the ground. The whole antenna could be hidden by a 6 foot privacy fence.
So far I have been discussing the 20 degree (electrical heighth) version and all the pictures are for the 20 degree antenna. I also built one for 30 degrees. It is 6 feet tall instead of 4. The horizontal wires are still 32 feet long.
Polarization is vertical,
thus the "vertical" in the name. The radiation from the horizontal wires cancels and the radiation
from the vertical wires reinforces. It radiates somewhat like a four-square vertical array except you can't control the pattern. Signals are strongest broadside with a null off the ends.
What's the difference between this antenna and a cubical quad? This antenna is a magnetic loop rather than a cubical quad because of it's small vertical dimensions. The four foot radiators are only about 1/18th wavelength. One rule of thumb of when antennas switch over from electric to magnetic is 1/8th wave.
Comparing this loop to a compact loop from AEA or MFJ, the larger size makes matching much easier and conductor losses easier to control. I also suspect that the larger capture area helps the performance of the larger loop.