eumaies
made their 6th forecast (view all):
Probability
Answer
0% (0%)
Less than or equal to 1349
94% (+5%)
Between 1350 and 1499, inclusive
5% (-5%)
Between 1500 and 1649, inclusive
0% (0%)
Between 1650 and 1799, inclusive
1% (0%)
More than or equal to 1800
Why do you think you're right?

every week with low numbers it gets a little more likely to come in under 1500

Files
Why might you be wrong?

Wouldn't take a big unexpected spike to exceed 1500.

Files
ctsats
made a comment:
But was the number of +23 cases last week a low one?
Files
eumaies
made a comment:
Maybe i'm confused. I see the data through 9/6/2025 as showing 2 cases int he past week.  Or if that data is incomplete 11 cases in the week prior to that.  Which data shows 23 cases in the last week?
Files
ctsats
made a comment:

You are not confused - there are indeed 2 different ways for measuring "new" cases: one is to check the number indicated in the last week, as you say here (which is indeed 2, and it has been even 0 in previous updates); that number will always (I mean, always) be revised upward in next updates, usually significantly: the previous week, now shown at 11, was at 0 when it was first reported, 2 weeks ago, and most probably it will be more than 11 at the Sep 10 report (and possibly even higher at the Sep 17 one)... Compare the reported 25 cases of the week starting Aug 17 in the latest update with the update of just 3 weeks ago (Aug 20), when this number was reported as 0 (well, it was obviously not zero)...

The other, safer way, is to count the aggregate difference between updates: so, 1431 total cases in the last update minus 1408 total cases in the previous one gives us 23 "new" cases (mind the quotes). This number and its evolution is actually what we are trying to guess here, and it is heavily influenced by revisions of previous weeks, not so much of the latest one. And this number has not been low in the last weeks (if we stick to the average of the last 4-5 weeks, i.e. ~20/week, we are easily above 1,500 total cases by Sep 30)... The fact that most people here are upping their estimates for bin 3 after the latest update is neither unreasonable nor irrational. 

Files
eumaies
made a comment:

Super helpful, thanks.  So the weekly numbers get revised upward, and typically within about 2 weeks of when they are first posted.  That's helpful, and I'll revise accordingly.



Files
ctsats
made a comment:

Revisions routinely go further than just 2 weeks in the past; and then again, we may have data bombs, i.e. bulk revisions of past data, such as the one we had in June: https://www.randforecastinginitiative.org/comments/174210

Files
Files
Tip: Mention someone by typing @username