*I'am the author of original post. And whole this situation makes me frustrated, some people started to rewrite my words with some misstakes, give not completely accurate answers for some reasonable questions and so on. Also general readers made some conclusions that I did not provide and so on.
>> Third, they performed no statistical tests.
We are not crazy, first of all we performed stat tests. We performed chow test for structural break at 17th nov (p-value about 1e-16) and chi-sq test (p-value ~1e-160).
>> First, the authors did not sample the population randomly, they surveyed the population and requested self-reported data.
Only plarium can sample population randomly and operate with fround truth data. We can just try to deal with self reported.
For example we made subsample with people who started filling drop before potential change and got same results for them as for whole dataset. We tried to find any changes in raw data (a lot of new users, average fill rate changes and so on), that can be sign of data corruption (covid in your example), but we did not find anything. (btw, I don't think that covid is reasonable example, but I understand what you mean).
Generally I have following point: you can select group of users (big enough) and collect their drop every day. It does not matter how did you sample them from general population if you assume that all people are equal (in our case - have same drop chances). If you see some prooved changes in theirs drop, you can claim that there is some changes in general population drop. I mean that if all people are equal all subsamples will have have same distribution as general population.
The only way when it is not true - people was not equal from the beginning. It can be because of many different reasons, for example - unannounced compensation system, when drop prob depends on your previous drop. But we have no reasons to think so, this theory was not prooved by any previous research. Personally I tried to find something like month guarantee for sacred shards (or sacred shard + leg book) (according to my feelings, you can't get less then 4-5 sacred shards from 2x top UNM chest per month) but failed, everything looked like fair random.
* I have to notice, that it is steel possible that we deal with some year-compensation (if you got to many sacred shards this year, you will get more epic books in last two month of the year???) but we don't have any data to speak about it. It is possible, but not very likely, I suppose.
Any way only plarium can say final word. And they assert that everything is okay.
I think this situation has two potential finals (if it is bug or a/b test):
1) plarium will rejet/fix it, we will see it in our data and report that now drop is as usual
2) plarium will accept changes, apply it for all users and we will see it in datamined data. (oh, I have so many questions to plarium, why they were generating drop on client side :D )
Alternative (I don't want to belive in it) - it is not a/b test or bug, plarium divided users by some rule (like they do with donate offers) and we will not see any changes in our data or datamined one. This situation is a dead end and we will never come to agreement.
So lets just wait...
Good luck :)