Hey Klose,
There is no hard and fast rule for which tool to use and when. You can use a variation of tools if that suits the situation.
At this stage I would say keep going with the course, in chapter 4 Chunk demonstrates with finesse more about shading and sharp edges.
The new CORE | Fundamentals of Mesh Modeling course also covers this topic.
There was this old video that it's still relevant today of Jonathan Williamson comparing the results of Sharp, crease, weight, holding edges, etc and seeing which gave the best result. I can't find the video, but in essence, it's always best to use edge loops if you want the cleaner result. It'll cost you in geometry but if you want to do it by the book making edges loops is the way to go. The other methods, like creasing will give you muddier results, but it is more efficient geometry wise so it's best to use any of them in objects that are secondary and are going to be farthest away from the camera.
Like was mentioned previously, there isn't really any hard or fast rules for using any of the three options! If you can achieve the results you need using any of them, then the easiest option (or combination of) would be my preferred way to work.
However, I personally use two of them at certain stages consistently:
Bevel Weight Edges - During the high-poly stage, we generally want more edge loops at edges to create higher fidelity so we can bake that detail down. You could manually create those loops, as Omar mentions that is likely the cleanest method with the most control, but then you're juggling 3-5 vertices at every "point" to edit your mesh. I prefer to use a Bevel modifier and tag the edges I want to bevel/add more loops to. This way I can simply move one vert around, instead of multiple.
Sharp Edges - When I am creating a low-poly version of my model, chances are those edge loops I created for the high-poly are going to get cut. We do cover it a little later in the course like Adrian mentioned. The issue is that once those edges are gone, the shading suffers since we no longer have the required topology to shade it correctly. Instead, we can use "Sharp Edges" to tell Blender how to better shade the model, which will also improve the quality of our baked assets.
I personally don't really use the Creased method, my needs are met with the other two options but it is still nice to have in your toolkit just in case!
Understanding their uses will help you to develop your criteria.
So bevel weight is when you are using the bevel modifier set to weights If you're not using the bevel modifier then it does nothing. This allows you to adjust different beveling amounts based on the bevel modifiers settings. The criteria becomes: using bevel modifier, need variation on the amount of beveling throughout the mesh.
Crease is used to give sharp edges(points for vertex crease) when using subdivision surface modifier aka subsurf or sub-D modeling. You use crease where supporting loops won't work, aren't possible to make, and/or will create to much geometry.
Sharp edge/face is used to control the normals calculations in a way to counter act the overall mesh settings. For example if you have edges/faces that you want to render as flat when the mesh is set to smooth shading. This can also be used with smooth by angle(use to be auto smooth)for edges/faces that are in the smoothing angle.