My general feeling on the by election result

[andrewducker]

Reminder that any voting system where you can win a single seat without 50% of people thinking you're better than the alternatives is not fit for purpose.

Comments

[simont]

But that's the point, of the options that *could* be elected 50% of the voters preferred one option more than the other.

Which is why I said that you do end up with 50% of the voters preferring the winner to somebody – namely, to whoever survived until the last notional round and then lost it.

It's true that if the elimination process doesn't need to go the full distance then this might look more sensible. And of course I'm not arguing that bloody FPTP is better – when we had the choice to switch to this kind of system I voted in favour. But I don't like to see overstated arguments even for my own position (perhaps especially not for my own position), and I think it's disingenuous to describe it as "50% preferred A to all the alternatives" when what you really mean is "all the alternatives except the ones that my own algorithm decided didn't count".

[andrewducker]

But it's not to *somebody* it's to *everyone who might plausibly win*.

And I don't think it's fair to say "the ones that my own algorithm didn't win". We're talking about a single seat here. If we were talking about multi-member constituencies then which algorithm you use, how you divide up the seats between multiple parties, etc. makes a big difference. But when it comes to a single winner situation I don't think I've seen many alternatives seriously discussed which would leave the party that 7% of the voters had as their first choice in active competition. Are you thinking of an algorithm which wouldn't do that?

(I'm ignoring things like Approval Voting, because it doesn't have very wide support, and seems to leave itself open to strategic voting in a way that looks complex and unpleasant if it becomes common.)

[simont]

Ah, so when you said "preferred to the alternatives", your idea of which alternatives to consider was derived from some first-principles idea other than "the set of candidates eliminated during the IRV process"? That would make more sense, but in that case, was definitely unclear in your original statement of "any voting system that doesn't do this isn't fit for purpose".

Also, I think there are several plausible ways to define the set of not-hopeless candidates. The one that sprang to my mind was the Smith set. But apparently IRV can elect a candidate outside that set, by sometimes eliminating the Condorcet winner if there is one!

[andrewducker]

My idea is very basically - that nobody should be considered voted in, in a single seat, unless they are preferred by at least 50% of voters over the other options. I gave an example of how you might do that, but I am totally open to alternative methods of finding out what people prefer.

(Also, did you see people being nice about you over here: https://andrewducker.dreamwidth.org/4560569.html?thread=32304313#cmt32304313 ?)