Because of delay in reporting. Their reporting is very delayed unlike in most of the countries so for "yesterday" you have only a small fraction of real cases, for "week ago", about half, and all of them only for month and more ago...

The other countries aren't reporting real-time data either, they just don't report the date of death.

who would be the ones that are reporting closest to real-time?

I’d guess it’s those with very few deaths. New Zealand’s were mostly reported on the day. However some were identified as covid related further down the track, delaying reporting.

It's really not that serious compared the black death in the 1300's

I think that if you have to compare a disease to 1300's it is pretty serious

So what is a good or best practices solution to this? I was thinking you could branch the graph line 10 days prior into two lines, the top line being "projected", the bottom line being "actual", with labels, and maybe with a shaded area in the middle.

Honestly, for data intended to be viewed by the lay public, the best option is probably to show by day reported. It's a little laggy, and will blurs things a bit, but realistically you're probably most interested in the moving average anyway, so more lag and blurring isn't that much of a problem.

Showing a graph always going down is just poor public communication; people will inevitably misconstrue it.

Here you go: https://adamaltmejd.se/covid/

this is great, thanks for the pointer, great example for effective communication

Awesome job you are a hero

Perfect

related: in .at there currently is a debate re whether the lockdown measures are working or not and we see similar mistakes made by journalists, politicians and pundits on social media who naively compare incomplete data from the current week to the previous week