A Real World Opportunity for Comparative Voting System Analysis!
-
@ArendPeter This sounds like a perfect fit for the new beta site.
-
Yep, we're working on a new version of star.vote (currently live at dev.star.vote).
For comparing voting methods, one option is to setup a multi-race election where each race votes for the same set of logos but uses a different voting method.
Here's the ones we support (some of them aren't enabled at the moment, but we could enable them if needed)
- STAR
- Ranked Robin
- IRV
- Approval
- Choose One
- Score
-
Are you going to just use one method, or are the counter & the voters willing to vote & count by several voting systens?
My arguments for Approval are most important for public political elections. But of course, given that I claim that Approval is the best choice for public political elections, that, of itself, is a good reason to use Approval.
But, if demonstrating Approval because it's good for public political elections isn't the main consideration, then I'd suggest a rank-method.
The big advantage of Condorcet (I'd use RP(wv) ) over Hare is that Hare potentially has a lesser-evil sucker-voting problem, wherein some people bury their favorite under a lesser-evil. But I don't expect that problem to obtain in your election.
I've conducted 3 rank-balloting elections where an actual collective choice had to be made. In all 3 instances, I used MinMax(wv),
In the 1st one, we were voting on which of 3 district-sizes to propose in a PR proposal. Someone said, "Let's use Pairwise [Condorcet]." Sure, so I conducted a MinMax(wv) vote. Of course the middle CW won. It wasn't my favorite, but at least it wasn't the worst choice. But the two advocates of the opposite non-CW choice rejected that middle CW, & walked out of the meeting, ending any progress there.
In the 2nd poll, our local LWV group was deciding on what voting-system our local chapter would endorse. Everyone but me preferrred IRV, because it was what they'd been propagandized with. I offered Approval & Condorcet. My own favorite was probably Condorcet We discussed for a while, & then I suggested that we vote. Given that Condorcet was the most strategy-free, & my favorite, I suggested Condorcet, & they agreed. I use MinMax(wv).
Approval won, as the middle CW, Everyone but me rejected that result.
That's two Condorcet votes in which the CW was rejected.
In the 3rd poll, I & my two sisters, & the boyfriend of one of them were at a house in a big city, & we decided to go to a movie. I probably went out to get a newespaper, if we didn't already have one. I suggested that we all look at the movies playing, & then nominate our favorites among them, & then vote by Condorcet, which I explained to them is a rank-balloting method.
The CW was something that I didn't like. It was BORING !!
The others seemed to like the idea of going to that one. I don't remember whether it was a middle CW or a majority CW. ...only that I didn't like it.
Therefore, for non-political votes, or especially fun things like Pizza topping or choice of a movie, or for artistic choices like your vote, Hare would be my choice, unless it's desired to demonstrate Approval for public political voting.
Here's my brief definition & instruction for Hare:
Repeatedly, eliminate the alternative that tops fewest ballots, till something tops most of them., It wins.
Hare has another advantage for small-room handcounts: Its handcount is just as easy as Approval. Typically approximately the same number of votes need to be counted.
Condorcet of course becomes much more time-consuming if there are more than about 3, or maybe 4, alternatives. With 3 alternative there are 6 vote-totals to count. With 4 altenatives there are 24, which I'd say is too maniy to expect the other peope to be willing to wait for.
But, if you're going to use Condorcet, use the best one: Ranked-Pairs(wv):
If no one is unbeaten, due to one or more cycles:
Drop the weakest defeat in each cycle.
The resulting unbeaten alternative wins.
In a vote with few voters, there could be a pairwise tie, resulting in a vote tie. If so, I suggest eliminating all but the tie alternatives, & re-apply RP(wv) to them. ...repeating until a single alternative wins.
If the process doesn't converge to a single winner, then just choose the alternative with the most top-rankings.
If you want a Condorcet-Criterion method with a faster handcount, use Sequential-Pairwise(Borda). That means that you first count the rankings by Borda (defined below). Then arrange the alternatives in a vertical list with higher Borda scoriing candidates BELOW lower scoring ones.
Let me define Borda:
In the version I'd prefer, each ranking gives to each candidate a number of points equal to the number of alternatives not ranked above it,
(Alternatives that a ballot doesn't include in its ranking are regarded as equal-bottom ranked on that ballot.)
Alternatively, instead of "not ranked over", it could instead be "ranked below". I prefer the former.
Anyway, you've got that vertical list of alternatives. Among the top two, find out which one pairbeats the other. i.e. find which one is ranked over the other on the most rankings. Eliminate the pair-loser. Then do the same between the winner & the next alternative down the list. Keep doing that, A pairwise count between the current top-2 in the list, & delete the pair-loser, Eventually only one alternative will remain. It wins.
That's for a Condorcet count that's quicker than the usual exhaustive pairwise-count. ...& it always chooses from the Smith-set (the smallest set of alternatives that all pairbeat everything outside that set).
But I don't really know why you'd choose Condorcet for this count,
Here are my suggestions, in order of preferenfe:
- Approval (if it's desired to demonstrate the best public political method)
Otherwise:
-
Hare (if you want the favorite candidate among a mutual-majority instead of a possibly little-liked middle CW).
-
RP(wv) if there are only 3 alternatives, or if a computer count is being done.
-
SP(Borda),, if you want a faster Condorcet count.
-
@michaelossipoff we used a flexible ballot that could be converted into any of the used systems. It was a score ballot with a stated approval cutoff.
There were 13 possible voters and 12 candidates. 7 people submitted ballots. We used STAR, Approval, Score, Condorcet, and Choose-One. Half interesting and half boring, all of them produced the same winner. I don’t really count the results of choose-one in this context as a legitimate instance, since the conflicts of interest are essentially removed by knowledge that other systems will also evaluate the data.
I also tried out the “symmetric quantile normalized score voting” (SQNV) method, which produced the same winner.
-
@cfrank said in A Real World Opportunity for Comparative Voting System Analysis!:
@michaelossipoff we used a flexible ballot that could be converted into any of the used systems. It was a score ballot with a stated approval cutoff.
There were 13 possible voters and 12 candidates. 7 people submitted ballots.
You’re talking about some earlier poll. I was asking about the current poll whose initially-stated ending-time is midnight tonight (whenever it occurs where you are).
Quantile? Is that like quintile & quartile?
You didn’t say what or who the candidates were, or what or who won.
Anyway, I was asking about the current poll, with evidently 3 participants. As I said it’s declared end of balloting is midnight tonight.
We used STAR, Approval, Score, Condorcet, and Choose-One. Half interesting and half boring, all of them produced the same winner. I don’t really count the results of choose-one in this context as a legitimate instance, since the conflicts of interest are essentially removed by knowledge that other systems will also evaluate the data.
I also tried out the “symmetric quantile normalized score voting” (SQNV) method, which produced the same winner.
-
@michaelossipoff oh I see, I was referring to the original context of this post, which was a low-stakes github icon contest for the lab I work in lol. People submitted 12 candidate icons. I'm not sure what the polling is, but if it's voting systems, my polling is given by my "sign off" below my posts.
SQNV is the system referenced here:
https://www.votingtheory.org/forum/topic/470/symmetric-quantile-normalized-score
also here: github/SQNV
-
I'm not sure what the polling is,
It starts here and is taken up here.
In general about keeping up with posts to this forum, the initial page of the forum, as pointed to by the home page of the website, shows the posts that were most recently posted or most recently received comments. It may be necessary to refresh it.
but if it's voting systems, my polling is given by my "sign off" below my posts.
symmetric-quantile-normalized-score [10] cardinal-condorcet [9] ranked-condorcet [9] approval [8] cardinal-metric [7.5] ranked-bucklin [7] star [6] ranked-irv [5] ranked-borda [4] score [3] for-against [2] distribute [1] choose-one [0]
There are three problems with this list relative to the poll described above (which has reached its deadline anyway). 0) you mention voting systems that have not been nominated, 1) you vote in a system that has not been nominated, and 2) you didn't post it in response to the discussion of the poll, at the second reference above.
-
@jack-waugh I take you don’t perhaps intend to come across as aggressively combative toward me, but I am not easily receiving your messages with another tone. If you have something constructive to say, by all means have at it. Pettiness is not something that will be appreciated on this forum.
-
@cfrank, I look forward to your publication of a tally that includes your vote and others'.
-
@jack-waugh I don’t follow.
