Enough people are asking that I thought I'd explain in one thread my take on the recent EDGES paper on a possible detection of the global 21cm signal. nature.com/articles/doi:1…. 
I've done a couple of media interviews: insidescience.org/news/astronome… and sciencemag.org/news/2018/02/s…, but they don't get into the details.
Firstly, let me say that I have immense respect for the EDGES team. They've been consistently careful and methodical. They've sat on this result for a long time, cross checking it. But, I'm a bit skeptical of their claim of detecting a -0.5K absorption feature at 78 MHz (z=18)
My concern is that the spectral feature in question may be systematic imprinted by their instrument. The beam of their telescope changes with frequency and can pick up more or less of the bright synchrotron emission from our galaxy.
If there is a little bit of wobble in the size of the beam versus frequency, it can imprint that same wobble on the measured spectrum, and that wobble can wind up as the residual plotted in Figure 1b.
So the first question is, does their model of their antenna produce anything like this wobble? The answer, plotted in Figure 4c, is yes. The width of their beam wobbles, and at about 40-deg off of zenith (dashed line) it has almost the exact same shape as the residual.
So their beam *could* affect the residual. Does it actually? Well, their choice of beam model seems to have a strong effect on the amplitude of the residual, as shown in Table 1, in the 4 rows below the horizontal line. Even the choice of modeling software (FEKO or HFSS) matters.
If the beam is imprinting this signal as a systematic, it should depend on where bright features on the sky fall within the beam. We should see some time dependence of the fit signal. Table 2 shows that the fit amplitude does indeed vary substantially over the course of a day.
All of this suggests to me that this signal is not passing some of the consistency checks I'd expect it to. This doesn't mean it's wrong, but it does mean that it looks possible to me that some or all of the signal they detect could have been caused by instrumental systematics.
Add on to this the fact that this signal, at -0.5K, is larger than the max signal we could have expected from gas cooling adiabatically after decoupling from CMB. Given a known CMB and gas temp, the max 21cm signal should have been -0.3K. Getting to -0.5K requires new physics.
Extraordinary claims require extraordinary evidence. For an experiment that must control systematics at the 3e-5 level to make a first detection, I think it is much too early to speculate about new physics. The focus should be on corroboration & deeper analysis.
