The progress so far:
The actual glass material:
Result:
A fresnel mirror for the model:
Result:
Combined approach:
First, let’s understand how glass material works in real life. Being an almost completely transparent material, it’s not opacity that makes it seen. Here’s a reference image to figure it out:
http://www.homeshopcatalogue.co.uk/catalogue/images/Rock-Glass-E049.jpg
As we can see, the albedo color and opacity have a minimal role here and can be safely ignored. We see glass via these factors:
Refraction
Surface specularity
Surface receives shadows
I handled the refraction on my setup the standard way, might go more in depth later though:
Seeing how the translucent blend mode didn’t handle the results I needed, I tried the masked blend mode and fed a fresnel into the input. Here’s how that turned out:
(fyi, the Fresnel phenomenon is that non-metallic surfaces turn into a perfect mirror when viewed from an angle close to 180 degrees). So basically this is a step in the right direction though causes specularity to affect the shadows. And the material still lacks “substance”, and the masked blend mode doesn’t support refraction.
So I decided to combine both approaches - use two instances of the same mesh with different materials - one to handle the specularity, one - the refraction and opacity. Doesn’t look much, but definitely better:
I’ll keep you updated on my progress, and whether I find a better approach than placing two instances of the same mesh at the same location - which is a bit “dirty”, if you ask me.
Good Work! Love this!
Here’s the full shader network used to calculate glass specularity via Fresnel effect:
As you can see, I used the xnormal-baked AO map to emphasize the shape of the glass - and it’s an useful tool to manipulate reflections.
Thanks mate 
I saw many people had a question with this. Just trying to help and share thoughts.
There is word that Translucent specularity is handled in 4.8; but how it is done - it’s yet to be seen.
Sorry, had wrong image for shader network; fixed it now.
Both meshes side by side, to compare: the fresnel reflections are handled by the one on the left, and the standard glass material handling the refraction is on the right.
The combined approach yields better results; but still far from perfection I’m afraid.
Remember to set the Fresnel mesh not to cast shadows; or else reflective parts will affect dynamic shadows, which isn’t exactly how you’d expect your material to work. Self-shadowing only is another possible choice.
Updated the first post to reflect the progress so far.
After some fine-tuning, here’s what my materials started to look like - a lot more acceptable and more surface, though now it has a bit of “milkiness” when exposed to light directly.
Try using the Advanced Lighting Map from the starter content, glass relies on reflections and transparency to look good, and this starter map is a bit too bare.
I want it to look good in any lighting condition 
that’s why I’m using this map. It’s about getting the physics right. But thanks for the advice.
When I finish the shading I’ll try to get the shadows and caustics.
Before rolling in the next version, here’s a bit of advice - DO NOT BAKE AO MAPS. Generate them from the normal map. Because the inside of the model is going to look awful. Also, a curvature map would be fun to play with, to map the reflections.
