[Dylan16807]

> Approval voting is the range compressed version of score voting.

I'm not thrilled with score voting to begin with, and compressing it down doesn't change much.

> Use score voting and you can give your favorite candidate a different score than your second favorite without giving them the same score as your least favorite.

That doesn't solve my problem. Take exactly what I said before, but replace "check" with "give the maximum score to".

> Both of them are still better than RCV.

I don't see how.

[ClayShentrup]

score voting is extremely good.

https://www.rangevoting.org/BayRegsFig

compressing it to approval voting doesn't hurt very much and makes the ballot extremely simple.

> I don't see how.

i just showed you how. better average voter satisfaction, better resistance to strategy, etc.

https://www.rangevoting.org/BayRegsFig

not to mention radically simpler:

https://web.archive.org/web/20190219005733/

https://sites.google.com/a/electology.org/www/approval-votin...

IRV is a strategic mess that maintains a two party system.

wonk.blog/duopoly

https://clayshentrup.medium.com/later-no-harm-72c44e145510

https://clayshentrup.medium.com/star-voting-is-simpler-than-...

[Dylan16807]

> i just showed you how.

That comes across as weirdly aggressive when you're referring to earlier in the same post. Like I'm dumb for not somehow seeing that before I finished my own post.

Also what's the definition of honesty for approval voting? Where am I supposed to draw the line?

The chart is pretty but I'm not trusting that chart without a lot more info.

> IRV is a strategic mess

I didn't say I want instant runoff in particular.

[ClayShentrup]

it doesn't really matter how you define "honest". what matters is how people actually vote. the only really crucial distinction for "honest" versus "strategic" is that "honest" can't consider viability. in the simulations, we just use "approve everyone you like more than average". optimal strategy is to approve everyone you like more than the expected utility of the winner. https://rangevoting.net/RVstrat6

here's lots of detail. you can run the code for yourself. https://www.rangevoting.org/BayRegDum.html

here's a slightly different model from the late harvard stats phd, jameson quinn. https://electionscience.github.io/vse-sim/

don't "trust" anything. look at the data. you could also read the book "gaming the vote" by william poundstone.

[AnthonyMouse]

> I'm not thrilled with score voting to begin with, and compressing it down doesn't change much.

Compressing it makes it slightly worse, but some people think it makes it easier to understand. I tend to think that's silly; people can understand "score each candidate on a scale of 1 to 10" perfectly well. But approval voting would still be a significant improvement over FPTP, and even over RCV.

> That doesn't solve my problem. Take exactly what I said before, but replace "check" with "give the maximum score to".

RCV doesn't solve that either, because it's subject to Arrow's Impossibility Theorem. It's actually even worse, because it makes you give a lower rank (rather than an equal one) to your most favored candidate to prevent an even worse candidate from winning.

Suppose your favorite candidate is the first choice of 20% of the population, your second favorite is 25% and the two candidates you hate are at 25% and 30% respectively. RCV makes you give your top pick to the second candidate so they can beat one of the two candidates you hate and ensure the runoff isn't between both of the candidates you hate.

Meanwhile with score voting your favorite candidate might have won, because they were the first choice of only 20% but the second choice of everyone else, and then end up with an average score of e.g. 6 when all the others are at 4 and 5.

RCV tends to do the opposite of that. If you have two opposite extremists and a moderate, the moderate can get knocked out in the initial round and then you get a coin flip between the two opposite extremes, even if the moderate would win one-on-one against either of the extremes. Or, in a district that skews to one side, give the district to the other side, because the minority party makes it to the runoff with 40% of the vote, the majority party splits between two candidates each at ~30%, but then if the majority party's extremist goes to the runoff instead of its moderate, they could lose the runoff to the minority party's moderate and give the district to the minority party.

[Dylan16807]

> Suppose your favorite candidate is the first choice of 20% of the population, your second favorite is 25% and the two candidates you hate are at 25% and 30% respectively. RCV makes you give your top pick to the second candidate so they can beat one of the two candidates you hate and ensure the runoff isn't between both of the candidates you hate.

Let's call those candidates A B X Y.

I don't see the issue. I vote my actual ranking, A is eliminated. If this scenario isn't super weirdly convoluted, more of those votes shift to B than X, so X is eliminated next. Now it's B versus Y. If I vote for B instead, the same thing happens and we also get B versus Y.

And what happens here in score voting could be basically anything. Not enough specifics.

> If you have two opposite extremists and a moderate, the moderate can get knocked out in the initial round and then you get a coin flip between the two opposite extremes, even if the moderate would win one-on-one against either of the extremes.

Yeah that's a real issue, and I'd want to use condorcet methods to fix that.

> Or, in a district that skews to one side, give the district to the other side, because the minority party makes it to the runoff with 40% of the vote, the majority party splits between two candidates each at ~30%, but then if the majority party's extremist goes to the runoff instead of its moderate, they could lose the runoff to the minority party's moderate and give the district to the minority party.

Meanwhile with approval or score voting, a lot of people pretend not to like the other party's moderate because they're trying to get their side to win and it gets really messy. Is there a need to have multiple candidates per party? If there is, a two stage election that picks party first and then picks probably-the-moderate is probably better than anything.

[ClayShentrup]

> I vote my actual ranking

no you don't. take a warren supporter who votes biden to stop trump. that person ranks:

biden 1st, warren 2nd, trump 3rd

they don't want warren to be a spoiler.

http://rangevoting.net/TarrIrv

[Dylan16807]

That person is not me. Don't tell me what I do.

[ClayShentrup]

that person is STRATEGIC VOTERS, who are about 90% of the electorate. even if you want to vote honestly, you're still subject to the whims of other voters.

indeed, practically the only thing that matters in a voting method is the aggregate affect of what it achieves in light of other voters. YOUR vote will almost never make a difference. you could upgrade voting methods in exchange for giving up your right to vote, and you'd still be drastically better off.

[Dylan16807]

> that person is STRATEGIC VOTERS

When I said "I" I meant me. This one person. And I was talking about that specific scenario.

It wasn't a statement about "strategic voters" in other scenarios.

[AnthonyMouse]

> If this scenario isn't super weirdly convoluted, more of those votes shift to B than X, so X is eliminated next.

It doesn't have to be that convoluted, all it takes is for the eliminated candidate to be a moderate so their votes go in two different directions. But you're right that I messed up that example; the percentages are wrong.

The problem case is when your second most favored candidate would otherwise be eliminated first and you need to prevent that by causing your most favored candidate to be eliminated instead, because the second best candidate has a better chance in the next round.

Suppose the candidates you dislike, X and Y, are the first choice of 40% and 25% of people respectively, and then A and B split the remainder evenly. X and Y are the two extremists -- on opposite sides of each other, with the moderates A and B in the middle. You favor A but A leans in the direction of X and B leans in the direction of Y.

If B is eliminated first then half of B's support goes A but half goes to Y, Y is still ahead of A and then A is eliminated next. If A -- your preferred candidate -- is eliminated first, half their support goes to B and the other half to X but Y gets nothing. Y then loses to B and the final round is X vs. B rather than X vs. Y. And the elimination of Y puts all their support behind B since X is the opposite extreme. But only if you rank B above A even though that's not what you'd have preferred.

> Yeah that's a real issue, and I'd want to use condorcet methods to fix that.

But now you're no longer using RCV/IRV. Score voting is a Condorcet method.

> Meanwhile with approval or score voting, a lot of people pretend not to like the other party's moderate because they're trying to get their side to win and it gets really messy.

Except that doesn't really help them because doing that also makes it more likely that their least favorite candidate wins, which is a significant incentive not to do it. The only reason to do that is if you're confident your favored candidate could only lose to your second choice, in which case it was really a two candidate race to begin with.

> Is there a need to have multiple candidates per party? If there is, a two stage election that picks party first and then picks probably-the-moderate is probably better than anything.

It doesn't matter if they're in the same party or not. If you're using a voting system that allows more than two parties to be viable then you'll have similar candidates running from similar parties. "Force the election to be one candidate from each of two major parties" is FPTP and it's terrible.

[Dylan16807]

> A B X Y

It's very unlikely that the two candidates I hate are on opposite extremes and both popular.

> Score voting is a Condorcet method.

That gains it a point but there are much better methods. Ones where I don't feel the need to balance risk versus reward for candidates I moderately dislike. Then the guarantee of the pairwise winner coming out on top is actually using accurate information on what everyone wants.

> Except that doesn't really help them because doing that also makes it more likely that their least favorite candidate wins, which is a significant incentive not to do it. The only reason to do that is if you're confident your favored candidate could only lose to your second choice, in which case it was really a two candidate race to begin with.

If I'm moderately confident then I'm likely to do it despite the risk.

> It doesn't matter if they're in the same party or not.

The whole framing of the problem was that one of the candidates in the minority party wins. If there are four unrelated candidates that problem goes away. The more popular moderate won, not a big deal.

[ClayShentrup]

> That gains it a point but there are much better methods.

utterly false.

https://electionscience.github.io/vse-sim/vse-graph.html

[AnthonyMouse]

> It's very unlikely that the two candidates I hate are on opposite extremes and both popular.

It's easy to hate the candidate on the opposite extreme from the way you lean, so all this really requires is for the extremist on your side to be a corrupt populist who gets support by telling people the lies they want to hear or is paying off the right people to get favorable media coverage or valuable endorsements. Or is just more extreme than you can accept but you're in a district with some people who want that.

Notice also that neither of these candidates are the first choice of the majority. They just have enough support in a >2 candidate race to not be the first knocked out.

> Ones where I don't feel the need to balance risk versus reward for candidates I moderately dislike.

This is Arrow's Impossibility Theorem again. All of them do that, because in the rock-paper-scissors triangle where no candidate can beat both of the others, you then need something equivalent to a score to choose the winner. At which point degrading your second choice hurts them against both your first and third choice and whether or not you should do that is influenced by how likely you regard it that other voters will favor your first choice over your third but not your second.

It's also a dangerous game because the error bars on polls are huge and it's more often than not that the final results are very different than anybody's wild guess from the day before they started counting the votes.

> If I'm moderately confident then I'm likely to do it despite the risk.

Suppose your true ratings would be 10 for your most preferred candidate, 7 for the second best and 1 for the inhuman monster the opposing tribe somehow supports for no explicable reason. Polls say your first choice is expected to score ~5, your second choice ~6 and your vile enemy ~4, but all of these are plus or minus 2 points or more because polls are practically random number generators. What are you going to do?

You have the option to try to tank your second choice to give your first choice a better chance, but it's still a very real possibility that your first choice ends up at 4 and the hated enemy at 5.

> The whole framing of the problem was that one of the candidates in the minority party wins. If there are four unrelated candidates that problem goes away. The more popular moderate won, not a big deal.

It's still the same, and the minority party candidate isn't necessarily that much of a moderate, they're just not a far extremist.

Suppose it's California and a Republican, a Democrat and a member of the Green Party are in the race. The district is 40% Republican and under the old system correspondingly 60% Democrat, but in a system with more than two viable parties, it's 40% Republican, 29.9% Democrat and 30.1% Green.

If you hold that election with RCV and the Democrat gets knocked out first, the moderate Democrats (which, with the Green candidate in the race, was all of them) have to choose between a California Republican and a Green Party candidate who proudly wants to raise the gas tax to $8/gallon and pull out of NATO. More than a third of the moderate Democrats choose the Republican over that and under RCV that becomes a Republican seat.

The same race with score voting only does that if people vote the way you seem to think they would, which is exactly their incentive not to.

[Dylan16807]

> This is Arrow's Impossibility Theorem again. All of them do that

They do not. There are voting methods where I feel they work well enough that I don't worry about strategy, and I don't worry about what threshold to use for approval, I just rank choices honestly and then I'm done.

> Suppose your true ratings would be 10 for your most preferred candidate, 7 for the second best and 1 for the inhuman monster the opposing tribe somehow supports for no explicable reason. Polls say your first choice is expected to score ~5, your second choice ~6 and your vile enemy ~4, but all of these are plus or minus 2 points or more because polls are practically random number generators. What are you going to do?

In this situation the strategy play is that I rate both of the first two at 10. I'm not risking #3 by being strategic, I'm risking that I get #2 when I could have gotten #1.

For the situation where I might strategically lower the score, it's when there's a 10, there's a 7, there's a 4, and there's a 1. Do I give the 4 an honest rating that might help them win over the candidates I like, or do I give them a 1 and pray really hard that neither of the two guys I hate win? It's a tough choice.

> Suppose it's California and a Republican, a Democrat and a member of the Green Party are in the race.

This is a totally different situation. In the last version there were two moderates and one of them won. It was fine. In this version there's one middle candidate and they lose. But we already did this scenario earlier. I said "Yeah that's a real issue, and I'd want to use condorcet methods to fix that." So not just an instant runoff, but keeping the ranking system.