fizzy said:
But CMB is an incredibly noisy measurement and results derived therefrom seem to be crude fits to a sparse number of data points huge error bars.
It's the complete opposite. Because the CMB is so incredibly bright, and because the physics that are relevant to the production of the CMB are so simple, it's an incredibly accurate probe of certain features of our universe.
In this specific instance, as of the latest Planck results the CMB determines the average dark matter density of our universe to within about 2%.
For example, most astronomical observations are plagued by the fact that there's a lot that we're
not seeing: do we not see certain galaxies because they aren't there, or because they were too dim for detection? Incidentally, this is what motivated the news post in the OP: a model of how many galaxies are there that are too dim to see.
The CMB doesn't have this problem, because it has nearly uniform brightness in every direction on the sky. There's no worry that there might be parts of the CMB we're not seeing. Plus, at certain frequencies, the small differences in temperature from place to place on the sky are brighter than anything else across more than 90% of the sky. To get very accurate measurements, we can use measurements of the CMB across a range of frequencies to subtract these foreground signals, in addition to just not using data from the small area where the foregrounds are really bright (e.g. the center of our galaxy, radio quasars).
The physics that produced the CMB are also really simple: you can use simple integrals to model what the temperature differences should be (a modern computer can calculate the predicted statistical properties of the CMB given a model to very high accuracy within just a couple of seconds). This is as opposed to galaxy formation, which is a monstrously complicated phenomenon that is often modeled using complex N-body simulations and lots of questionable approximations.
