Why does the ray fire from the cage's edge along the average normal?

Hi nnaiad ,

I don't know this, but it might have something to do with the fact that with split Normals, a lot of rays would simply never hit the Object:

?

Or even worse, in the wrong spot?

@jlampel will certainly be able to answer this.

Although the Cage is not Smooth Shaded, the information of the averaged Normals is still available (or easily calculated).

The Rays are not dependent on the Normals of the Cage.

Like Spikey, I'm not entirely understanding of how this works.
I would suspect by enabling the cage option in the bake properties may alter the normals.

On further reading, the blender manual says,

When not using a cage the rays will conform to the mesh normals. This produces glitches on the edges, but it is a preferable method when baking into planes to avoid the need of adding extra loops around the edges.

 The inward rays are casted from a version of the active object with disabled Edge Split Modifiers.

The normals (rays) as I understand it, on the cage object the angles of the rays are dependent on position.
Something like this:

These are not exact angles but a guide

What I meant was:...not dependent on the 'Shading Normals' of the Cage. (The Cage doesn't even have Shading...)

I don't think you can control the ray direction manually, but again, I am not sure about all this.

Boy, that escalated quickly. I'm guessing the cage does to the normals its own thing under the hood, without you having to set shade smooth or anything. Blender be nice like that, saving you a hassle or two. 

nnaiad

don't worry about those angle values too much; the side of the Object has to be represented on the side of the Cage and because the Cage is larger, the corresponding points of the Object are in an angle to the Cage:

Cool in depth explanations. And even though we see that picture above as a cube from up top and a cage with the normals pointing at the cube, I like to think of it as a trampoline, with the elastic tendons supporting the thing where you jump at the middle.

It's amazing, how many complicated CG concepts can be explained by trampolines

Exactly as Spikey said, visualized as if the cage is extruded from the mesh, with many more rays than in the visuals.
Knowing how things work is useful, but try not to get bogged down by the details of how and why. 
It will probably have little or no effect on the result of your Blender project.

Martin's picture above is perfect! A cage only works because it has the exact same vertices as the low poly object. The bake can then use the imaginary line that's drawn in between each matching vertex to guide the rays INSTEAD of using the normals. 

That looks right to me, naiad.

Hi Thomas NNokota ,

If you use Smooth Shading, it changes the Shading Normals of the Mesh...indeed something like in my picture, but the Ray is cast like the green arrow and finds the Normal in that point to be like the blue arrow (should be straight down):

I hope that makes sense.

"Is smooth shading changing the way Blender calculates the Normals?"

Yes!

Have a look at the Blender 3.0 Basics Course: https://cgcookie.com/lessons/intro-to-blender-and-cg-cookie

This lesson for instance:  https://cgcookie.com/lessons/mesh-normals-and-shading