That data is ONS reported COVID19 deaths (not excess deaths). It will probably be updated based on the reclassification of 28d, which I don’t personally agree with, but standardisation is OK either way.
What I did was model observed deaths by day of the week using log-linear, quartic and cubic regressions for all weeks from 15 onwards (with an extra factor for day of the week). Then selected the model with the best overall goodness of fit – this was quartic from Week 22 onwards – which is what is plotted. You can see that the data has a periodicity, but the fluctuations increase with time (hint: it’s plotted on a log scale). Forward projection just increases the magnitude of fluctuations due to the quartic nature of the overall trend- it’s a simple model. I really wanted to understand whether we could report ZERO deaths because, as per Scotland, the media will get excited.
When ONS update their data, the extinction probability may change.
Excess deaths are back to historic ten-year range. Forget the noise about being BELOW. they are as expected at the moment. A few age groups and regions have a little excess. Nothing of concern though at the moment. Excess deaths is insensitive to changes because of the reporting lag by ONS. That’s why the daily PHE data is used on a daily basis.