I've been involved in this lengthy circular debate: What should be the autoconfirmed age and article count in the Hebrew Wikipedia? See https://phabricator.wikimedia.org/T243076 if you curious about this particular one, but I'd love to ask a more global question:

How were these numbers calculated originally?

For the account age, the default is four days, or five or seven days for a few wikis.

For the edit count, the default is zero, but several wikis have 5, 10, 25, or 50.

(See https://noc.wikimedia.org/conf/highlight.php?file=InitialiseSettings.php and search for "wgAutoConfirmAge" and "wgAutoConfirmCount".)

Some wikis have groups, usually called "extended confirmed", and with higher counts; for example, 500 edits in English and some other languages (search for wmgAutopromoteOnceonEdit on the same page).

So, how did the people arrive at these numbers? Why is it four days by default? Is it all just intuition and guesses, or was there any research behind it?

Is it *good* that four days is the default for everyone, until someone bothers to update it (most wikis don't)? Or is it just a coincidence that was defined for a certain wiki and applied elsewhere? And when it's updated, why is it updated to one number and not some other?

While I am an ardent supporter of the "anyone can edit" principle, it makes general sense to have some restrictions based on edit count, account age, and perhaps other parameters. But HOW are they calculated? Would it make sense to anyone to start making some calculations around it and optimize the number for wikis of different sizes?

I'd imagine that there could be a calculation that says "in a given wiki, the chance of being reverted or blocked goes down after X days and X edits", and this number is probably different for every wiki (maybe there already is such a calculation somewhere). This could possibly be a starting point for a good calculation of a threshold; it wouldn't be perfect, because in some wikis it can perpetuate community practices which may be biased against new editors, but at least it's based on data and not on guesses.