@Jgalt Is it just me or do all countries keep saying that and then not deliver?

Because the question was scheduled to be open for most of 2020, the early resolution in this instance caps all the possible points at a mere fraction of the total. The points are optimized to provide the incentive to always predict your best credence, and I think it works for all instances. (Though it starts to fail if someone cares more about having a lot of points in 1 month rather than 1 year.) If you had gotten the entire total of points for the rest of the question duration, this scoring rule could incentivize people to predict really high early o...


Indonesia has recorded another sharp rise in coronavirus cases, with 227 people now recorded as infected. That 55 new cases since Tuesday, the biggest single-day rise the country has recorded to date.

More significantly, the death toll has nearly quadrupled to 19 people, up from five.

(Source: https://www.smh.com.au/national/coronavirus-u…)

I think this must be >15k infections, possibly a lot more.

— edited by ninjaedit

I just realized that there are some ways in which airbnb is really useful during times like these. Stranded foreigners need places to stay once hotels close, and shutting down airbnb would be bad from this perspective because they might then have to stay somewhere more crowded. I don't see a strong need to crack down on airbnb from as social distancing perspective. It's rather the opposite.

— edited by ninjaedit

@pranomostro The US already has 14,000 confirmed cases and I'm hoping that the WHO report will catch up to at least that number in the next 7 days.

But I guess there's a chance that I'm missing something. The US has different types of testing going on and maybe the WHO report only counts the ones done officially by the CDC or some other national body? So there's some chance the numbers are diverging for reasons other than delay. I think delay is most likely, though.

@(Jgalt) I'm worried he may have a point. Italy's naive CFR is at 8%. Spain is already at 5% too. China's naive CFR was at 2.2% for a long time and climbed to 4% when case growth slowed down. So there's a good chance Italy and Spain's naive CFRs will continue to climb. Are Spain and Italy testing fewer of the cases that don't require hospitalization than China did? That seems somewhat implausible to me because I don't think China tested many of those milder cases either, and I also heard a report of one Italian doctors going from household to household ...

@DavidMathers This is the first time I'm below the community prediction on anything covid-19 related. Starting to worry that it might have something to do with being located in the UK. :P

@Tamay Suggestion: CFR should be higher in cases where treatment dynamics change because of the size of the outbreak. 4.6M infections worldwide probably won't put an unmaneagable strain on hospitals, but 769M certainly would. In the former scenario, 0.5-1% CFR seems more appropriate to me, whereas in the latter scenario 2% might be a bit too low even, unfortunately.

— edited by ninjaedit

@Jgalt Arguably large parts of Germany are still safer than some parts of the US (esp. when you look at the ratio of confirmed diagnoses to deaths, on which Germany is doing best in the world and the US worst or second worst in the world). I don't think the community is factoring in that Level 3 warnings would look a bit ridiculous if they go out to countries that have better prospects than the US.

@ninjaedit The evidence that points at IFR >0.3 seems really robust to me, so I'd want to see at least a >2 sentence long description about how sample bias was avoided before I take an antibody study seriously that suggests something even just in the vicinity of that.

but probably better than the -200 MIPs you’d otherwise pay.

Because the question will be open for really long, I think he'd lose <10 points even with a maximally off prediction. Similarly, whether I initially predict the peak for NY numbers to be 3 days ago or 15 days in the future doesn't make much of a difference, as long as I update my prediction once we know the actual result. (This is why I think it's valuable to think more carefully about resolution timelines and the size of the range.)

If the antibody test becomes available later this year, I wonder if there are practical advantages to taking it. Maybe having a certificate that you have the antibodies could get you entry into other countries that wouldn't otherwise allow you entry.

@postmort Countries with good testing are all going into lockdown. The question is where the death ceiling will end up, not how fast the growth rate is. If you expect most countries to have been early enough to start their lockdown before a point where the situation gets worse than Wuhan, and if you expect th few outlier places that were too late to get more like 10k deaths instead of 30k, there's a chance this could resolve at <100k deaths.

— edited by ninjaedit


Someone should tell him that 0.15% of New Yorkers are already dead & test positive %age is nowhere near 50% there even among the people they test (which are heavily biased toward symptomatic and/or healthcare workers).

I'm starting to think that a peak in August is >20% likely. Or whatever the end of winter is for African countries or countries like Indonesia. It depends on thinking more about the resolution criteria (because reporting of confirmed cases may be poor in some areas), but unfortunately I'm becoming increasingly convinced that warmer temperatures don't completely slow the virus, and that containment may be hard to achieve in countries with weaker health systems, infrastructure or government preparations. I think the peaks we'll see in European countries an...

@notany I have a strong intuition that this is a poor reply. I don't think those factors make that large a difference. Do you really think you can get a factor of 3x from really crude demographics like that?

If yes, then that's a major consideration I'm currently not thinking about right.

(One thing that is interesting that deaths in Iceland have gone down again slightly.)

— edited by ninjaedit

India (1.38B), Indonesia (270M), Pakistan (220M), Nigeria (206M), Bangladesh (164M). I don't think lockdowns will likely work well in all those places. We also know heat does not slow down the virus enough to halt outbreaks (Jakarta's funerals were 1,100 higher than all-time high of 3,300 for March alone). And we know that IFR is 0.5%-0.9% for a context like the UK's situation (e.g., 0.15% of NY city is already dead and their test positive %age is <40%, and they primarily test healthcare workers and people with symptoms). So, for the countries I listed, ...
The website initially just shows the categories "violent crime" and "property crime." But when one clicks on the top-down menu, one can also add "All other crime." With all three categories the typical crime rate is more around 5000–5600 rather than 3000–3600. So I guess my initial estimates were way too low. Here's me hoping that lockdown reduces crime by 50% at least. But now I made my estimates larger because even though the website seems to be a bit sneaky, I think the question text is unambiguous here and it wants us to estimate *all* crime. (There ...

@PabloStafforini Probably several things, but I could imagine that for Germany, the number of confirmed diagnoses is very close to the number of people who actually have the virus, whereas in France and Spain, the true number might be a multiple of it? I know that Germany has done a lot of testing. I don't know what the testing situation is like in France or Spain, but I could imagine that it's not unusually thorough.

@ninjaedit It's important to think about whether the IFR for countries like Indonesia might be higher or lower than the IFR for countries such as Italy or Spain. That's where my main uncertainty rests currently. I expect >80% of deaths to happen in non-European, non-US countries.