Thinking of the two "voter" criteria...
Somebody PMed me pointing out an example of someone included in DT1 only because they were trusted by a permabanned member. I was aware that permabanned members were allowed to be voters, but I thought it'd be too rare to matter. But now I'm leaning toward excluding them.

Also, I've increasingly been feeling that something should be done about inactive voters. Currently, a user can be inactive for years, but still contribute toward those two criteria. The specific way I'm leaning toward addressing this is to require that voters must have received merit from at least 2 distinct users in the last 3 years.

I did a pretend reshuffle, and compared to the 136 users who were eligible for DT1 last time, with the above 2 changes, these users were no longer eligible: bavicrypto, be.open, Best_Change, comit, digicoinuser, ekiller, ezeminer, finaleshot2016, Gunthar, Harkorede, Heisenberg_Hunter, hybridsole, joniboini, Koal-84, Lachrymose, mandown, and witcher_sense. (Some of these users may have become ineligible for reasons other than the above two changes; I just compared the output today after the changes to the output on Feb 1.)

If instead it was "voters must have received merit from at least 3 distinct users in the last 3 years", then these users would also become ineligible: Baofeng, MinoRaiola, and Russlenat. I'm leaning toward 2, mainly just on the principle of preferring a larger list generally.

What do people think of these two changes?

While I'm here, here's a thought that's been floating around in my head for a while: A while ago I realized is that if you are among the ~136 people eligible for DT1, then your probability of being in DT (DT1 or DT2) after the cross-DT1 exclusions are applied is the CDF of the hypergeometric distribution, with the shape of the function largely dependent on [the number of people eligible for DT1 trusting you minus those excluding you] (ie. your "net DT1 inclusions"). If your net DT1 inclusions is very positive, then you will almost always be in DT; if your net DT1 inclusions is very negative, then you will almost never be in DT; if it's within a few of zero, then it'll switch randomly back and forth. This got me thinking that maybe it'd be more elegant to determine either DT1 or perhaps even all of DT by just directly computing a probability distribution based on net-inclusions (perhaps with other inputs) for every user, and then shuffling according to these distributions every month. But it's just a vague idea in my mind at this point; I don't plan to do anything with this soon, and I might never pursue it.

~hypergeometric distribution~