<div dir="auto">Both Benham-Score and Score based Sequential Pairwise Elimination can be cast as Pairwise Sorted Score methods.<div dir="auto"><br></div><div dir="auto">In both cases you start with the candidates listed in Score order, high to low, top to bottom, and swap adjacent candidates that are out of order pairwise until the final list is a pairwise "beat path" finish order. All of the out-of-order adjacent pairs have been "rectified."</div><div dir="auto"><br></div><div dir="auto">The difference in the two methods is the rectification priority. For Benham, as long as there remains any adjacent pair out of order, rectify the pair closest to the top of the list. For SPE, as long as there is any adjacent pair out of order, rectify the pair closest to the bottom of the list. Thus we make a distinction between bubble sort and sink sort.</div><div dir="auto"><br></div><div dir="auto">In either case the Smith set will end up at the top of the list.</div><div dir="auto"><br></div><div dir="auto">However, if we think of rectification as correction of a tentative order, based on error prone estimated merit, why not start with the pair for which the score order is most tentative ... least confident?</div><div dir="auto"><br></div><div dir="auto">Suppose the tentative order looks like</div><div dir="auto">A*****************B**C, with B and C close in estimated score. Which adjacent pair appears to be more secure, i.e. less tentative in its order? Thinking in terms of hypothesis testing, which hypothesis would yield higher confidence A>B or B>C?</div><div dir="auto"><br></div><div dir="auto">That is the rationale for "Score Sorted Margins": prioritize rectifications with least confidence in their score order ... i.e. where their positions are only marginally distinguished by score.</div><div dir="auto"><br></div><div dir="auto">This sorting rule makes use of information that is ignored by the bubble and sink sort rules, and so gives it an edge (no matter how slight) on accuracy.</div><div dir="auto"><br></div><div dir="auto">It also satisfies a strong Reverse Symmetry property not enjoyed by other sorts of sorting: if every ballot score x is replaced by 100-x, the finish order will be precisely reversed.</div><div dir="auto"><br></div><div dir="auto">For these reasons, all else being equal, I currently recommend Score Sorted Margins over the other potential score/pairwise hybrids.</div><div dir="auto"><br></div><div dir="auto">Here's an analogy that might help explain it:</div><div dir="auto"><br></div><div dir="auto">Suppose that we want to have a Round Robin basketball tournament to rank the teams from best to least.</div><div dir="auto"><br></div><div dir="auto">It is agreed that if one team defeats all of the other teams, it will be the champion team. But what if no team beats all of the others? Then the teams tied for the most wins will have a free-throw contest to resolve the tie.</div><div dir="auto"><br></div><div dir="auto">But that just decides which team is ranked first. How about second, third etc?</div><div dir="auto"><br></div><div dir="auto">Somebody comes up with the idea of having the free throw contest moved up front to the beginning of the tornament, to serve not only its tie breaking purpose (should the need arise) but also to effectively and fairly seed the match order.</div><div dir="auto"><br></div><div dir="auto"><span style="font-family:sans-serif">Each team has each of their 12 players make as many free throws in a row as possible before missing. The teams are listed in order of total team successful free-throws, their team "seed score".</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">We start by interspersing question marks between the adjacent teams on the list.</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">Then as long as there remains one or more question marks, find the question mark separating the two teams closest in their team seed scores. Have these teams play each other. If the one with the higher seed score beats (or ties) the one with the lower seed score, simply remove the question mark.</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">Otherwise, swap the positions of the two teams in the list, remove the question mark between them, and place question marks between them and their other neighbors (if not already in place) to reflect the fact that their relative rank order has not yet been confirmed.</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">When all of the question marks are gone, the final team finish order has been reached.</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><font face="sans-serif">Why is contest priority given to the teams closest in seed score? </font></div><div dir="auto"><font face="sans-serif"><br></font></div><div dir="auto"><font face="sans-serif">Because if two teams are far apart in seed score, it is less likely that a game between them would result in a swap, i.e. it is less likely that they are out of order, i.e. more likely that just leaving them alone would not be a mistake ... </font><span style="font-family:sans-serif">so not as urgent to test them against each other.</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">Thanks!</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">-Forest </span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><font face="sans-serif">P.S. IRV is like having a tournament consisting of successive free for all games, where the team making the fewest points is eliminated from the next free for all. until there are only two teams left, at which point a regulation game by standard rules determines the tournament winner ... in all more like a demolition derby than a sports tournament.</font></div><div dir="auto"><br></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><br></div><div dir="auto"><br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">El mar., 7 de jun. de 2022 4:29 p. m., Andy Dienes <<a href="mailto:andydienes@gmail.com">andydienes@gmail.com</a>> escribió:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Hi Forest,<div><br></div><div>If I am understanding RSTAR correctly, it should be the same as electing "the lowest score candidate that defeats all of the higher score candidates," right?</div><div><br></div><div>It certainly sounds intuitive. How would it compare to, e.g. Score Chain Climbing?</div><div><br></div><div>-Andy</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Jun 7, 2022 at 6:36 PM Forest Simmons <<a href="mailto:forest.simmons21@gmail.com" target="_blank" rel="noreferrer">forest.simmons21@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="auto">Benham-Range is ISDA (department of agriculture?). In fact, it elects the lowest score candidate that defeats all of the higher score candidates. <div dir="auto"><br></div><div dir="auto">So it is a simple proposal with lots of heuristic rationales and criteria compliances going for it.</div><div dir="auto"><br></div><div dir="auto">Another similar ISDA proposal that I call RSTAR for reasons soon apparent, always elects the highest score Smith candidate in the three candidate case:</div><div dir="auto"><br></div><div dir="auto">RSTAR like STAR can be formulated as a procedure that takes a set of score ballots as input, and selects as output a winner from among the scored candidates on the input ballots.</div><div dir="auto"><br></div><div dir="auto">For comparison, here's a formulation of STAR in that format:</div><div dir="auto"><br></div><div dir="auto">Let beta be a set of Score ballots. </div><div dir="auto"><br></div><div dir="auto">Then the STAR output STAR(beta) is the winner of the two candidate runoff between X=Score(beta) and Y=Score(beta'), where beta' is the set of ballots with candidate X temporarily removed.</div><div dir="auto"><br></div><div dir="auto"><span style="font-family:sans-serif">The RSTAR output RSTAR(beta) is the winner of the two candidate runoff between X=Score(beta) and Y=RSTAR(beta'), where beta' is the set of ballots with candidate X temporarily removed.</span><br></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">See the comparison (and tiny contrast)? If you see it and understand it, then you know that RSTAR must stand for Recursive STAR.</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">This is the most succinct formulation of SPE(Score), Sequential Pairwise Elimination with a Score based agenda.</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">Since SPE is a respectable, classical method with a hoary history of use in deliberative assemblies, as prescribed by Robert's Rules of Order, this would be an impeccable public proposal.</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">The Recursive formulation should serve only as a bridge between STAR and the less exotic SPE formulation; we don't want any heads to explode!</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">What say ye? </span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">-Forest</span></div><div dir="auto"><span style="font-family:sans-serif"><br></span></div><div dir="auto"><span style="font-family:sans-serif">["Ye" is an archaic form of "ya'll" or "youse guys" ;>)]</span></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">El mar., 7 de jun. de 2022 10:51 a. m., Forest Simmons <<a href="mailto:forest.simmons21@gmail.com" target="_blank" rel="noreferrer">forest.simmons21@gmail.com</a>> escribió:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="auto"><div>Very good observations ...and I like your method suggestions. Benham-Range (without the renormaliization),unlike ordinal Benham, should also be monotone and precinct summable if I understand it correctly.<br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">El mar., 7 de jun. de 2022 5:12 a. m., Kristofer Munsterhjelm <<a href="mailto:km_elmet@t-online.de" rel="noreferrer noreferrer" target="_blank">km_elmet@t-online.de</a>> escribió:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On 07.06.2022 05:49, Forest Simmons wrote:<br>
> VSE (voter satisfaction efficiency) simulations seem to bear out that<br>
> STAR is a significant improvement over plain old Score voting, but not<br>
> quite as good as Score restricted to the Smith Set.<br>
> <br>
> So it appears that simple STAR is the low hanging fruit worth some trial<br>
> and error tweaking experiments to convert it into the best public proposal.<br>
<br>
I'd say there are two ideas of generalizing STAR, call them cardinal and<br>
ordinal. The cardinal one should preserve the property that in an LCR<br>
election where the centrist is highly rated, he wins, but where he's<br>
ranked poorly, he loses. (Making this cloneproof without just degrading<br>
to plain Range is hard!)<br></blockquote></div></div><div dir="auto"><br></div><div dir="auto">That's the idea of my suggestion ... pit the Martin Harper Lottery top probability candidate against the score winner ... or I could have suggested a runoff between the two top probability lottery candidates. </div><div dir="auto"><br></div><div dir="auto">Better yet ... do Sequential Pairwise Elimination on the support of the MH Lottery, where the agenda order is from least to greatest probability.</div><div dir="auto"><br></div><div dir="auto">The MH Lottery is based on an idea for</div><div dir="auto">Range-->Plurality DSV: given all of the Range ballots, how should you decide which candidate ballot B should support (from the superstitious one-person-one-vote point of view)?</div><div dir="auto"><br></div><div dir="auto">Answer: the candidate X should have a high score B(X) on ballot B, and should have high support from the other ballots as well. The support of X from the other candidates can be gauged by its total score S(X). So X should be the candidate that maximizes the product B(X)S(X).</div><div dir="auto"><br></div><div dir="auto">Which is why the Martin Harper Lottery probability for candidate X is the percentage of the ballots B for which X maximizes the product B(X)S(X).</div><div dir="auto"><br></div><div dir="auto">[In 2001 Martin Harper posted to the EM list a definitive explanation for diehard IRV supporters of how Approval utterly satisfies "one person one vote" by way of a memorable Maxwell's Demon vote shuttling mechanism. So it is only fitting that Harper get the credit for this lottery idea!]</div><div dir="auto"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
The ordinal one, on the other hand, should find the CW whenever he<br>
exists, and probably also elect from some nice set (Smith, Landau or<br>
Banks). STAR itself doesn't do this.<br>
<br>
> Some brainstorming is definitely in order. Ted Stern has been working on<br>
> this.<br>
> <br>
> Some possible directions:<br>
> <br>
> 1. Simplify the description of Score restricted to Smith to be on a par<br>
> with the simplest description of STAR<br>
> <br>
> 2. Find a better runoff opponent for the score winner.<br>
> <br>
> 3. Compare STAR with Score Sorted Margins and Sequential Pairwise<br>
> Elimination based on a Score agenda.<br>
> <br>
> Here's one idea:<br>
> <br>
> For each ballot B, and each candidate X, let B(X) be ballot B's score<br>
> for candidate X. Let S(X) be the sum over B of B(X). Then the score<br>
> winner is the candidate X that maximizes S(X). Elect the winner of the<br>
> runoff between the score winner and the candidate X that on the greatest<br>
> number of ballots B, maximizes the product B(X)S(X).<br>
<br>
I haven't thought about it in detail, but I think Range's de jure IIA<br>
compliance implies that Benham-Range and BTR-Range are both Smith. They<br>
might even be equivalent to Smith,Range. So if you want to look for a<br>
simpler way of explaining Smith,Range, those may be it.<br>
<br>
My thinking is like this, for Benham-Range:<br>
<br>
Benham-Range eliminates the current Range loser until there's a CW.<br>
Since Range passes IIA, eliminating without normalization doesn't change<br>
the order of the other candidates. Hence all but one Smith set member<br>
will be eliminated, after which the highest scoring Smith set member<br>
becomes the CW and wins. So Benham-Range is just Smith,Range and may be<br>
easier to explain.<br>
<br>
Benham-BTR I'm much less sure about because it protects the loser who<br>
pairwise defeats the other. This is more like STAR. It might be that a<br>
prospective loser is protected all the way up the Smith set, and then<br>
beats the highest rated Smith set member pairwise -- unless such a<br>
scenario is impossible.<br>
<br>
For runoffs, I think you could argue in two ways. Either the runoff is<br>
about telling two quite good candidates apart, or it's about giving each<br>
faction of the electorate "their" candidate to vote for. The former<br>
suggests (if it's a manual runoff) just picking the two Smith set<br>
members with the highest rating; the latter, something more like "the<br>
candidate rated highest by voters who rated the first candidate low".<br>
<br>
If I had to choose, I'd probably go for the former, though it could<br>
affect the turnout in the general: if both candidates are good, a lot of<br>
voters may simply decide not to vote in the general.<br>
<br>
Automating the former, STAR would probably do a good job because it's<br>
unusual for a candidate to be the CW yet not in the top two by highest<br>
rating. At least it gets the bang for the buck award, even if it's not<br>
perfect :-)<br>
<br>
-km<br>
</blockquote></div></div></div>
</blockquote></div>
----<br>
Election-Methods mailing list - see <a href="https://electorama.com/em" rel="noreferrer noreferrer" target="_blank">https://electorama.com/em</a> for list info<br>
</blockquote></div>
</blockquote></div>