Ironically, if you visit this link with Javascript disabled, it shows: “AI scrapers break the web, to use this page you’ll need JavaScript enabled.”
The irony hurts and that wont stop scrapers but it will be shitty for some older screen readers
The disappointment is real, guess I’m going to stay on 13 for a long time, thank fuck for LTS!
Why does it appear so complicated? What are all these calculations?
Debian votes using the Schulze Method, which is a Condorcet method..
Condorcet methods attempt to answer the question “who would win if we asked everyone to vote on every pair of candidates”. Most of the variations between methods deal vary only where there are cyclic preferences e.g. people vote B over A, A over C and C over B.
They’re broadly considered the fairest voting systems available and they permit a candidate that everyone thinks is okay to win over polarising candidates (unlike first past the post and IRV, but like approval voting).
Back to the numbers, the big table is the reconstruction of the pairwise votes (people actually vote by ranking candidates but the ordering is taken to show your pairwise preferences)
Below that is the Schulze process steps being carried out. At the bottom the “Schwartz set” is the set of top, cyclic candidates. There is no cycle so there is only one candidate remaining and therefore a Condorcet candidate wins (the one people in general vote over everyone else).
Gotcha, how does this compare to true preferential voting?
Australia’s system is IRV (for the federal government / lower house). Schulze is significantly better.
IRV has the tendency to eliminate moderate candidates in the first round (only first preferences count in the first round) which is often where Condorcet candidates lie.
IRV doesn’t enforce two parties like FPTP but it still promotes polarisation.
