My preference assumes speakers with a particular set of radiation pattern characteristics, and also assumes a particular set of priorities. I wouldn't try it with most speakers, but with suitable speakers, I'd probably try it in most rooms.
There are four reasons why I prefer to set up for 15-20 degees off-axis, but again ONLY if we're using speakers which work well at that angle, which I think includes speakers like the 8083A, D&D 8c, and D&D 15c:
The first reason is, I like to use time-intensity trading to get a good soundstage across an unusually wide sweet-spot region. So when saying "I prefer to listen 15-20 degrees off axis", I'm envisioning something like 45 degrees of toe-in, and maybe more, with the speaker axes criss-crossing in front of the listening area. I can explain time-intensity trading if you're not familiar with it.
Second, I like to have a lot of time delay between the direct sound and the strong onset of horizontal-plane reflections. The above mentioned strong toe-in results in very weak same-side-wall first reflections, along with much later-onset and fairly strong opposite-side-wall first reflections.
Third, with an axisymmetric horn or waveguide, there is often a dip in the on-axis sound at high frequencies. This is caused by the mouth reflection, which arrives after a short time delay due to its slightly longer path length than the straight-down-the-middle sound. If we are directly on-axis, that mouth reflection arrives coherently. And there will be a frequency at which the mouth reflection arrives 1/2 wavelength later thatn the straight-down-the-middle sound, resulting in cancellation at the frequency.
Let me show you. Here are the off-axis curves for the Dutch & Dutch 8c, credit to Erin's Audio Corner. Focus on what's happening around 12 kHz on-axis and off-axis:
View attachment 530601
As you can see, there is an on-axis dip from ballpark 11 kHz to 14 kHz.
Now look at the orange curve, which is 20 degrees off-axis. Twenty degrees is far enough off-axis that the mouth reflection no longer arrives coherently. And the net result is arguably the best-looking curve. So imo it would make sense to make
that your actual listening axis, unless something else (like maybe 15 degrees) is even better.
And remember that this is a comb filter effect, so that dippage is not present in the reflection field. Also, its exact frequency changes somewhat as the listening (or microphone) distance changes.
And fourth, I like to minimize the spectral discrepancy between the direct sound and the reflection field. In general the treble will be "hottest" directly on-axis, so by listening from 15-20 degrees off-axis, we are sending the hottest treble into the reflection field rather than directly to our ears, which would have maximized the aforementioned discrepancy.
So yeah deliberately listening at 15-20 degrees may seem like madness at first glance, but there is method in it.
I think it's a 3-way (compression driver; 15" front-firing midbass; twin 15" rear-firing subwoofers). However the compression driver COULD THEORETICALLY BE a coaxial unit, in which case it would be a 4-way.
