Ranked Robin Disadvantages -
-
@masiarek Ranked Robin is a ranked-ballot method so it's expressive in that voters can distinguish between all candidates, but a voter can't say by how much, so that's where it lacks expressiveness.
My main problem is that for a Condorcet method, it's not a very good one. Obviously it's been kept simple, but that's not the only thing that matters.
First of all, as far as I can see it's just the Copeland method but for some reason rebranded as Ranked Robin.
But anyway, it fails independence of clones (like STAR voting actually). Lots of Condorcet methods pass this, so it's not because of mutually exclusive criteria or anything like that. So I see this as a poor failure.
But also, it works by seeing which candidate wins the most head-to-heads. But if there isn't a Condorcet winner, then in a real-life election, there's probably going to be a three-way cycle, and all these candidates will have the same number of head-to-head wins, so it will almost always go to the next tie-break mechanism rather than the method reliably producing a winner without arbitrary add-ons.
I always see Ranked Pairs as the default starting point for Condorcet. It's simple and has good criterion compliance. So that's what you need to beat if you want to use an alternative method. I don't see Ranked Robin as doing this.
-
A few things, all.
@masiarek, if I rank candidate A first, candidate B second, and candidate C third, then I know the order I prefer each, but not by how much compared to each other. If I had a 5-star ballot, I might give A 5 stars, B 4 stars, and C 0 stars; or I might give A 5 stars, B 1 star, and C 0 stars. There’s no way for you to determine how I really feel about B based only on my rankings. However, if I start with a 5-star ballot and give A 5 stars, B 3 stars, and C 0 stars, then you know for sure that I would rank A first, B second, and C third. You can always extract a full set of rankings from a set of scores, but usually you cannot extract a full set of scores from a set of rankings. Scores contain strictly more information than ranks; therefore, a score ballot allows voters to express more information than a rank ballot, i.e. score ballots are more expressive.
In my opinion, a rank ballot that allows equal ranks allows for a sufficient degree of expression to consistently determine the candidate closest to the center of public opinion, and the simulations support my claim as good Condorcet methods perform on par with good score methods.
@Toby-Pereira, you are overestimating the frequency of Condorcet cycles in real-world elections. As more people have studied the question, the estimates have gone down and down.
Additionally, Ranked Robin fails clone independence in the opposite direction of Choose One Voting. This means that to gain an advantage, a party would have to support entire campaigns of multiple candidates who are seen as identical by the electorate. This is so difficult in practice that it legitimately can be dismissed as a real concern. Candidates like to differentiate themselves from each other, electorates do not behave predictably, and campaigns are egregiously expensive. These difficulties are further amplified under a method like Ranked Robin that incentivizes candidates to appeal evenly to the entire electorate (see @Marcus-Ogren’s new paper on Candidate Incentive Distribution).
Also, if there’s not a Condorcet winner, then there are multiple scenarios more likely than a top-3 cycle that Ranked Robin resolves simply. Check out the electowiki.
https://electowiki.org/wiki/Ranked_Robin
Furthermore, your claim that Ranked Pairs is simple is…absurd. I canvass for STAR Voting, a far simpler method, every day, and it truly is at the limit of what we can expect lay voters in America to digest.
Simplicity actually is the most important factor for a Condorcet method because…it’s a Condorcet method. By that metric alone, it excels at both accuracy and honesty, and is also sufficiently expressive.
-
@sass I appreciate your perspective. However, if the power is concentrated in parties rather than in individual candidates, then election packing is a perfectly valid concern for methods that fail independence of clones. We can’t judge a voting system by the economic landscape reinforced by a different system, it needs to be recognized that market forces will lead to adaptation. I think, personally, that independence of clones is a far more important criterion than participation, for example, since for the most part voter activity is individual whereas party and media activity is more centralized. I would like to see how the members of this forum rank the various criteria in importance (in terms of having a “high probability” of compliance).
I think something like ranked pairs is an excellent single winner system, even though it fails participation. I also think approval voting is excellent, even though it fails the Condorcet criterion (but gets close in a sense). On the whole, my thinking is that proportional representation with some clever system of checks and balances to mitigate block formation would be even more excellent.
-
@Sass, thank you for your post.
@sass said in Ranked Robin Disadvantages -:
A few things, all.
@masiarek, if I rank candidate A first, candidate B second, and candidate C third, then I know the order I prefer each, but not by how much compared to each other. If I had a 5-star ballot, I might give A 5 stars, B 4 stars, and C 0 stars; or I might give A 5 stars, B 1 star, and C 0 stars. There’s no way for you to determine how I really feel about B based only on my rankings. However, if I start with a 5-star ballot and give A 5 stars, B 3 stars, and C 0 stars, then you know for sure that I would rank A first, B second, and C third. You can always extract a full set of rankings from a set of scores, but usually you cannot extract a full set of scores from a set of rankings. Scores contain strictly more information than ranks; therefore, a score ballot allows voters to express more information than a rank ballot, i.e. score ballots are more expressive.
This only applies if the score range allows for it. A STAR 0 to 5 ballot does not allow for a full ranking of candidates if there are more than a handful standing, so it's not really very expressive.
@Toby-Pereira, you are overestimating the frequency of Condorcet cycles in real-world elections. As more people have studied the question, the estimates have gone down and down.
Additionally, Ranked Robin fails clone independence in the opposite direction of Choose One Voting. This means that to gain an advantage, a party would have to support entire campaigns of multiple candidates who are seen as identical by the electorate. This is so difficult in practice that it legitimately can be dismissed as a real concern. Candidates like to differentiate themselves from each other, electorates do not behave predictably, and campaigns are egregiously expensive. These difficulties are further amplified under a method like Ranked Robin that incentivizes candidates to appeal evenly to the entire electorate (see @Marcus-Ogren’s new paper on Candidate Incentive Distribution).
A party could field two candidates, or if that's not allowed, then essentially get round it by having a B party that's basically the same. These candidates would not want to differentiate themselves from each other because they'd be there as part of a grand plan.
In any case, it's not just about intentional cloning. It's like how IRV fails monotonicity. An IRV proponent might say that it's too hard to vote strategically to take advantage of this, but that doesn't matter. It might affect things anyway. Ranked Robin's lack of clone independence could affect a result, intentionally or not.
Also, if there’s not a Condorcet winner, then there are multiple scenarios more likely than a top-3 cycle that Ranked Robin resolves simply. Check out the electowiki.
I'm not sure exactly what I'm looking at. It's quite a long article. What's more likely than a top-3 cycle when there isn't a Condorcet winner?
Furthermore, your claim that Ranked Pairs is simple is…absurd. I canvass for STAR Voting, a far simpler method, every day, and it truly is at the limit of what we can expect lay voters in America to digest.
Simplicity actually is the most important factor for a Condorcet method because…it’s a Condorcet method. By that metric alone, it excels at both accuracy and honesty, and is also sufficiently expressive.
Ranked Robin may well be simpler than Ranked Pairs, but I still think Ranked Pairs is a relatively simple method. Plus, the way Ranked Robin is presented on that wiki article, it doesn't look simple. Those 8 levels of when there is a tie - what is going on there? It's horrific.
Also, other than Ranked Pairs, there's also Benham's Method. This is just elect the Condorcet winner if there is one. Then eliminate the candidate with the fewest top ranks and carry on until the is a Condorcet winner. Like IRV except the winning condition is being the Condorcet winner rather than having >50% of the votes. This is fairly simple too and cloneproof.
Or something like Smith//Score.
-
@cfrank said in Ranked Robin Disadvantages -:
@sass I appreciate your perspective. However, if the power is concentrated in parties rather than in individual candidates, then election packing is a perfectly valid concern for methods that fail independence of clones. We can’t judge a voting system by the economic landscape reinforced by a different system, it needs to be recognized that market forces will lead to adaptation. I think, personally, that independence of clones is a far more important criterion than participation, for example, since for the most part voter activity is individual whereas party and media activity is more centralized. I would like to see how the members of this forum rank the various criteria in importance (in terms of having a “high probability” of compliance).
I think something like ranked pairs is an excellent single winner system, even though it fails participation. I also think approval voting is excellent, even though it fails the Condorcet criterion (but gets close in a sense). On the whole, my thinking is that proportional representation with some clever system of checks and balances to mitigate block formation would be even more excellent.
On criteria, I would say that participation is very important, or at least it would be if it wasn't incompatible with so many methods. But this is one reason why I like score and approval - they both pass it. What else does? Borda - well that's awful. Then you have Descending Solid Coalitions. and Descending Acquiescing Coalitions. Oh and First Past the Post. Of them, I would say only score and approval are reasonable methods.
But generally, monotonicity and independence of clones are "cheap" enough that I think there's little excuse not to have them.
-
As I've thought about it more, if there's a Condorcet Winner, then cloning is irrelevant under Ranked Robin, making it an unreliable strategy.
Also, basically all Condorcet methods fail Participation. It comes with the territory.
Moreover, focusing on pass/fail criteria is the issue that caused voting enthusiasts not to achieve real-world progress for 200 years. The question is not "Does this method pass this criterion 100% of the time?"; the question is "How well does this voting method perform on this metric in practice?". Considering that cloning is only helpful under Ranked Robin when there's no Condorcet Winner and that scaled elections without Condorcet Winners are incredibly rare and difficult to predict, I see it as a nonissue.
And just to set the record straight, I think Approval and Score are great methods. I absolutely support them and would be very happy to see their use in public elections.
@Toby-Pereira I was on mobile, so the link didn't copy properly. Here's the section discussing frequency of ties:
https://electowiki.org/wiki/Ranked_Robin#Frequency_of_tiesI need to clean up the electowiki page, but the Equal Vote site on Ranked Robin is a much better reference:
https://www.equal.vote/ranked_robin -
@sass said in Ranked Robin Disadvantages -:
As I've thought about it more, if there's a Condorcet Winner, then cloning is irrelevant under Ranked Robin, making it an unreliable strategy.
It's not just about strategy though, as said above. A potentially "wrong" result could still happen by accident.
Also, basically all Condorcet methods fail Participation. It comes with the territory.
Yes, it's very hard for methods to pass it in general. So relative to other Condorcet methods this doesn't count against Ranked Robin.
Moreover, focusing on pass/fail criteria is the issue that caused voting enthusiasts not to achieve real-world progress for 200 years. The question is not "Does this method pass this criterion 100% of the time?"; the question is "How well does this voting method perform on this metric in practice?". Considering that cloning is only helpful under Ranked Robin when there's no Condorcet Winner and that scaled elections without Condorcet Winners are incredibly rare and difficult to predict, I see it as a nonissue.
I agree that overall performance (however one might measure it) isn't necessarily the same as just how many criterion boxes you can tick. However, if a method does fail a criterion, it still doesn't look good if there is another method that is as good elsewhere that also passes this criterion.
And just to set the record straight, I think Approval and Score are great methods. I absolutely support them and would be very happy to see their use in public elections.
That's good. I think they and Condorcet methods have merit.
@Toby-Pereira I was on mobile, so the link didn't copy properly. Here's the section discussing frequency of ties:
https://electowiki.org/wiki/Ranked_Robin#Frequency_of_tiesOK thanks, but I'm not seeing anything to suggest that a three-way cycle would not be the most common tie.
I need to clean up the electowiki page, but the Equal Vote site on Ranked Robin is a much better reference:
https://www.equal.vote/ranked_robinThanks for the reference.
-
This post is deleted! -
Natural, sincere, circular-ties are vanishingly-rare. When there is one, it will virtually always have 3 candidates,..meaning that Copeland (Ranked-Robin) will be indecisive every time there's a sincere top-cycle.
If there are lots of candidates, Copeland might often elect the CW, even in some strategic circular-ties, But, with only a few candidates, supporters of a candidate who beat nearly as many as the CW does have an easy win by insincerely voting several candidates (whom their candidate can pairbeat) ver the CW,
Also, of course Copeland depends on the vagaries of how various regions of issue-space are divided into separate candidates...something that has nothing whatsoever to do with any merit-justification for choosing a winner.
-
@michaelossipoff said in Ranked Robin Disadvantages -:
Natural, sincere, circular-ties are vanishingly-rare.
I actually want to say, we should plan for common cycles. That's not because they are common; it's because they might become common. Cycles are a metric of success, because they mean many similarly-popular candidates competing along non-ideological lines.
@michaelossipoff said in Ranked Robin Disadvantages -:
If there are lots of candidates, Copeland might often elect the CW, even in some strategic circular-ties, But, with only a few candidates, supporters of a candidate who beat nearly as many as the CW does have an easy win by insincerely voting several candidates (whom their candidate can pairbeat) ver the CW,
I actually think Copeland is a good choice, just because it's easier to explain than the Smith set.
But Ranked Robin having multiple stages actually makes it difficult to explain, IMO. Easier to explain Minimax: "eliminate candidates with the biggest losses." Actually, I think even River is easy to explain (Minmax ignoring cycles), but it's an A++ top-shelf method.
-
@lime said in Ranked Robin Disadvantages -:
@michaelossipoff said in Ranked Robin Disadvantages -:
Natural, sincere, circular-ties are vanishingly-rare.
I actually want to say, we should plan for common cycles. That's not because they are common; it's because they might become common. Cycles are a metric of success, because they mean many similarly-popular candidates competing along non-ideological lines.
True …& people have made reasonable arguments for why margins makes more sense than wv in a spontaneous circular tie.
But, even then, wv makes enough sense, is well justified.
…&, with its vulnerability to offensive strategy, margins shares IRV’s strategic quagmire, from the git-go.
@michaelossipoff said in Ranked Robin Disadvantages -:
If there are lots of candidates, Copeland might often elect the CW, even in some strategic circular-ties, But, with only a few candidates, supporters of a candidate who beat nearly as many as the CW does have an easy win by insincerely voting several candidates (whom their candidate can pairbeat) ver the CW,
I actually think Copeland is a good choice, just because it's easier to explain than the Smith set.
RP, while choosing from the Smith-set, doesn’t mention it in its definition.
Copeland has the abovestated justification problem & can invite & facilitate burial.
But Ranked Robin having multiple stages actually makes it difficult to explain.
Yes, & frequent need for a tiebreaker, because natural cycles typically have only 3 members.
[quote]
IMO. Easier to explain Minimax: "eliminate candidates with the biggest losses."
[/quote]Yes, but RP isn’t difficult to define, explan or justify.
[quote]
Actually, I think even River is easy to explain (Minmax ignoring cycles), but it's an A++ top-shelf method.
[/quote]River adds a clause to RP, & loses the autodeterence of wv RP & MinMax.
-
@michaelossipoff said in Ranked Robin Disadvantages -:
River adds a clause to RP, & loses the autodeterence of wv RP & MinMax.
Could you explain more? How does it lose autodeterrence? I was under the impression that River(wv) was effectively minmax with fewer spoilers.
