Variants of loops have been designed and promoted as having certain advantages, and one of those is the so-called figure 8 loop.
This article describes an NEC-4.2 model at 14MHz of an antenna similar to a commercial example.
The graphic shows the geometry. In this case the source is at the bottom of the lower loop, and the blue square is the tuning capacitor. The loop conductor is 22mm copper tube, the loop diameters are 1m, and the capacitor connection is 100mm wide. Commonly these are fed by a low loss auxiliary loop at the bottom of the lower loop, but the direct feed is quite fine for modelling the loop performance.
The antenna is quite affected by proximity and type of ground, the model locates the centre of the structure at 2m height (1m from ground to the bottom of the structure) above “average” ground. Capacitor Q is taken to be 1000.
Maximum gain (-1dBi) is at high elevation (52°), a result of proximity to ground, and it is nearly omni directional at very high elevation angles.
In fact, even at 30° elevation, there is not a deep pattern null, front to side is about 5dB.
Radiation efficiency in this configuration is 23%.
The antenna does not quite fit the criteria for a Small Transmitting Loop. Although the perimeter is 6.3m, the current amplitude and phase is more like a pair of 3.15m circumference loops in parallel in phase, a result of the shared connection to the tuning capacitor. So in some ways, it acts like a pair of co-phased parallel 1m diameter loops with their center displaced by the loop diameter.
It is compact in footprint, and a simple single loop using the same length of copper tube would performs fairly similarly, the simple loop performs significantly better than this figure 8 on 40m.
I might note that many of the construction pics I have seen show a vacuum cap but compromised by relatively high resistance connections, and it might not achieve the Q assumed by this model, and therefore have somewhat lower radiation efficiency.