News for hackers. Stuff that matters.
Re:Ranked-choice voting is on the ballot in New York City by thaumasiotes on Wednesday October 23, 2019 @02:55AM (#21331846)
I hate the way they've defined their metric:

> VSE is expressed as a percentage. A voting method which could read voters minds and always pick the candidate that would lead to the highest average happiness would have a VSE of 100%. A method which picked a candidate completely at random would have a VSE of 0%. In theory, VSEs down to negative 100% would be possible if a voting method did worse than a random pick

If VSE is going to be a percentage, it should be a percentage of the maximum happiness -- that is, the total happiness if everyone got their first choice, not the total happiness that is the highest any candidate can actually achieve. I say this mostly because it's much easier to calculate.

But more fundamentally, there is no reason to expect the worst-case candidate to be equally as bad as the best-case candidate is good. It's a huge mistake to rank systems on a scale from +100% to -100% where the units on the positive and negative "halves" of the scale are different!

And it's also a mistake to define 0% as whatever sortition achieves. It's trivial to construct a pathological system in which 0% VSE is equal to 100% VSE. (Everyone is equally un/happy with all candidates.) This should be impossible.

Re:Ranked-choice voting is on the ballot in New York City by thaumasiotes on Wednesday October 23, 2019 @03:52PM (#21338212)
Let's consider an example:

You have 100 candidates. For 99 of those candidates, the entire electorate agrees that any one of these interchangeable guys would do a fine job. Nobody prefers any of them to any other.

The last candidate is Mao Zedong. Under his leadership, the country's industrial and agricultural bases will be systematically destroyed, anyone who owns a rental property will be executed, and tens of millions of people will starve to death for no reason.

The electorate agrees that Mao is an undesirable choice.

So: any method that can't pick Mao achieves a VSE of 100%.

Sortition, by definition, achieves a VSE of 0%. But it's still pretty good -- there's a 99% chance of selecting one of the good guys, and a 1% chance of epic, country-shattering disaster.

A method that is guaranteed to select Mao has a VSE of, I assume, -100%. But it is much, much worse than sortition (the notional midpoint, 0%), while the VSE of 100% is only moderately better than sortition.

These numbers don't make sense.