So in this case I would like to see this info on Beyond3D list next time. What AA method game uses. MSAA or NOT (aside from temporal, qAA).
It
is MSAA in HS! It's just not resolving to the typical results.
Consider a conventional 2xAA on a white/black edge. White is (255,255,255) in RGB space. Black is (0,0,0). 2xMSAA would average the two producing mid-grey (128,128,128), which you'd see. In the case of HS, the intensity extends beyond standard (255,255,255) intensity. Thus if you have a super-bright white triangle from a sunlit surface, with a sample of equivalent intensity RGB (1024,1024,1024) (HS doesn't represent colours in terms of RGB components), and a second black sample of (0,0,0), the MSAA average will be equivalent to (512,512,512) which is beyond the dynamic range of RGB, resulting in a white average.
What you seem to be wanting is a more subjective analysis of AA levels. In an HDR game where MSAA cannot resolve higher intensities below white, you will get the equivalent of zero AA on some edges. However, where the dynamic contrast isn't beyond the resolving step, you'll get the full MSAA interpolation. How exactly could you qualify that in a technical analysis though? "This game is 4xMSAA, but sometimes looks 0xAA"? If you follow that route, you'd also have to say of other games, "this game has no AA, but looks like maybe 2xAA, even 4xAA in places." That's non-technical!
So here we just stick to the facts about the engine. Current hardware and engines tend to have AA faults with HDR that means you don't always get the full benefit of the MSAA steps even though it's implemented. C'est la vie!