[EM] Sims - method terminology

[Kevin Venzke]

Hi Leon,

--- En date de : Jeu 10.6.10, Leon Smith <leon.p.smith at gmail.com> a écrit :
> Two questions...
> 
> What do all the different abbreviations
> mean?   Not being a voting
> theory specialist,  sometimes I have a pretty good
> idea of what your
> abbreviation means,  other times I
> don't.   Are they semi-standard
> abbreviations that appear in literature somewhere?

Many of them probably not. I'll run through them quickly. I don't know
that I will define all the methods but one can always ask more questions
afterwards.

> > Method  BestC   WorstC  Top     Middle  Bottom
>  Dist    DistN
> > CdlASnc 91.8%   1.1%    87.3%   10.4%   2.2%  
>  52.978  1.306

Conditional Approval (sincere). This is an obscure method for which I
recently started using the term CdlA. This is intended to be a three-slot
method where the votes are counted repeatedly and voters add in their
second-slot preferences when the leader after any previous round was a
last-slot (i.e. non-)preference.

Sincerely voting this isn't very realistic and I would've just deleted
its results, except that it placed first here...

> > MMstrict        91.8%   1.1%    97.2%   2.6%
>    0.1%    53.010  1.314

This is strictly-ranked MinMax, a Condorcet method which in the three-
candidate case can be equivalent to Schulze, Ranked Pairs, and River.
I think this term is pretty well-known. In the three-candidate case,
when there is a cycle, remove the weakest win. (Which is unambiguous
without truncation.)

> > Bucklin 90.4%   1.4%    79.0%   10.2%   10.8%  
> 53.044  1.453

Bucklin is well-known on this list. In papers I have seen it reinvented
multiple times though. Woodall's QLTD is very similar. The voters
successively and simultaneously add more votes until a majority is reached
or all the votes have been cast.

> > DAC     89.6%   1.4%    61.8%   29.6%   8.7%  
>  53.049  1.481

Woodall's Descending Acquiescing Coalitions (DAC) method was published
in Voting Matters. In the three-candidate case it will elect the one
with the most first-preferences who has more first-preferences than
voters not voting for him (first or second). When there's no such candidate
the one with the most first and second preferences wins. (I usually just
call this "approval" to save words.)

> > MAP     91.8%   1.1%    94.1%   4.8%    1.2%
>    53.155  1.359

MAP is Majority Favorite//Antiplurality. I think only I use this term.
In this method we will check for a majority first preference and, if
there is none, we elect the candidate with the fewest last preferences.

> > RangeNS 83.3%   0.4%    46.4%   21.8%   31.8%  
> 53.237  2.083

Range, normalized sincere. In other words the voters will scale all their
ratings from 0-100 (or whatever) so that the best and worst candidates
in the race are at the extremes.

> > ApprPoll        81.2%   1.3%    52.0%   19.7%
>   26.9%   53.442  3.060

Approval following pre-election polls.

> > QR      84.1%   2.0%    33.2%   66.1%   0.7%
>    53.471  2.831

Quick Runoff. Elect the candidate with the most first preferences who
doesn't have a full majority pairwise loss to the very next candidate in
first preference descending order.

> > DSC     83.0%   1.5%    53.1%   40.5%   6.4%  
>  53.474  2.656

Woodall's Descending Solid Coalitions. In the three-candidate case this
will elect the candidate with the most first preferences among those
who have more first preferences than strict last preferences.

Note that DAC and DSC are identical if no voter truncates. DAC and DSC-tr
use truncation while DSC is here strictly ranked.

> > C//A    81.0%   1.8%    13.3%   78.9%   7.8%  
>  53.510  3.074

Condorcet//Approval. The double-slash has sometimes been used on this
list with the sense that the former method is used as a filter for the
latter. (With a double-slash, excluded candidates are completely 
eliminated before the second method is considered; Woodall uses a comma to
say that the original scores of the right-hand side method are used as a
tie-breaker for the method on the left.) So C//A is the method where if
we do not have a Condorcet winner, we elect the approval (first plus
second preferences) winner.

> > MMWV    81.0%   1.8%    17.3%   72.7%   10.1%
>   53.510  3.075

MinMax(wv) is MinMax with defeat strength measured as winning votes. The
strength of a pairwise win is equal to the number of voters who voted
for the winning side.

> > CdlA    82.5%   1.5%    25.1%   62.1%   12.8%
>   53.585  2.786

CdlA again, this time using truncation informed by polls.

> > ApprZIS 77.0%   0.9%    58.5%   13.2%   28.2%  
> 53.593  3.841

Approval with zero-information strategy; i.e. the voters will approve
every candidate who is better than average.

> > 2sMMPO  81.1%   1.3%    42.6%   28.6%   28.8%  
> 53.602  3.004

This is MMPO (see below) performed using approval ballots. (It uses the
polling information here.)

> > MMmarg  78.1%   3.0%    5.9%    62.4%   31.7%
>   53.762  3.983

MinMax(margins) is MinMax where defeat strength is measured as the
difference between the vote totals of the two sides of a contest.

> > IRV     79.1%   3.6%    1.1%    67.9%   31.0%
>   53.851  4.216

Instant Runoff Voting. Here we quite simply elect the pairwise winner of
the top two candidates (based on first preferences).

> > SPST    78.3%   2.5%    27.0%   44.1%   28.9%
>   53.996  4.245

Strongest Pair with Single Transfer (SPST) is a method I suggested three
years ago. (I realized recently it is misnamed, but I don't think it's
worth it to rename.) In this method, in the three-candidate case, we will
elect the first-preference winner unless he is the majority last 
preference, in which case we elect the pairwise winner of the other two
candidates.

> > MMPO    76.7%   4.4%    4.4%    34.8%   60.8%
>   54.017  4.877

MinMax(pairwise opposition). This is like MMWV except there's no concept
of a pairwise win. Elect the candidate with the fewest votes against
him in any contest regardless of whether he won that contest.

> > IRV-tr  76.3%   4.1%    0.1%    42.7%   57.2%
>   54.110  4.924

IRV but the voters will truncate according to the approval polling, for
whatever reason.

> > Raynaud 76.8%   4.4%    1.6%    34.4%   64.0%  
> 54.139  4.841

Raynaud with winning votes. In the three-candidate case, when there is a
Condorcet cycle, we will remove the win of the candidate who suffers the
worst pairwise loss.

> > QR-tr   76.0%   4.5%    0.1%    39.2%   60.7%
>   54.200  5.221

QR truncated following the polls for some reason. (I say "for some reason"
because these -tr methods satisfy Later-no-harm and technically don't
offer an incentive to truncate.)

> > VFA     73.3%   4.0%    11.3%   21.0%   67.6%
>   54.377  5.630

Vote For and Against. An old method by me. Elect the first-preference
winner unless he's the majority last preference, in which case elect the
second-place candidate.

Warren has called this "Venzke Disq Plur" I believe, which is a pretty
good name I think, as "VFA" seems to describe a ballot format more than a
method.

> > DSC-tr  71.8%   5.4%    13.1%   20.0%   66.9%  
> 54.796  6.735

DSC, truncated.

> > FPP     70.4%   8.2%    7.9%    12.8%   79.3%
>   55.249  8.487

First-Preference Plurality (sincere). Most first preferences wins.

> > Antip   44.8%   0.0%    8.8%    16.3%   74.9%
>   60.119  26.835

Antiplurality (sincere). Elect the one with the fewest last preferences.

------
 
> Are you willing to share your simulation code,  as
> open source
> software for people to review?

This is possible, but not immediately as I'm not sure it's finished. If
there are questions about how something is currently done I can show
the relevant parts of the code as needed, though.

Kevin Venzke