Hello, I recently played NFS U2 and noticed that the road reflections are quite accurate, compared to SSRR methods. It's mighty fast aswell, since I can play the game maxed out on a laptop, while SSRR is quite impossible on a laptop. I'm writing code(C#) and shaders in Unity 5 if thats important to point out.
So here is a video I found on youtube wich describes the reflections quite well:
My personal guess is that there is a small mix of screenspace rendering and a cubemap going on? Because I tryed a mirrored cubemap rendering method and the results were not as good. Infact they were terrible. I had 2 different results, one of where the reflected ground was hanging way beyond the reflection plane and the other had some normal issues, but was more accurate. Both required a cliping point input wich I calculated with a raycast under the camera. Not the best way to do it, but it worked for testing.
Here is a Unity3d link for the mirrored cubemap renderer method (not mine but the basics are the same):
http://twiik.net/projects/realtime-reflection-example-scene
Here are some screenshots I took from the video, on the left is the original and on the right is the mirrored view, for comparsion.
Another guess would be that there is some sort of a rendering matrix, and the rendered image is wisely placed along the surface?!
And exucuse me for my bad descriptions, english is not my natural language and my grammar might be of.