A compromise between Score Voting and Illusory Reform Voting

[NoIRV]

This probably has countless flaws, but:
(1) Each vote is a function f: candidates -> real numbers. At every moment it will be renormalized to the interval [0,1]. A vote is exhausted if all remaining candidates have the same score.
(2) Find the candidate with the smallest score and eliminate him. If he was given min or max score on a vote, that vote may need to be renormalized.
(3) Process continues to the last candidate standing.

Please analyze this system. I think it is immune to clones and prevents spoilers, but is not precinct summable and it is complex.

[William Waugh]

I didn't find any flaws.

I believe it exhibits formal balance, i. e. the balance of opposites. Say you vote for example Nader 1, Gore 0.1, Bush 0. I your opposite vote Bush 1, Gore 0.9, Nader 0. Say that because of the other votes, Nader is eliminated. Then your vote is inflated to Gore 1, Bush 0. My vote is inflated to Bush 1, Gore 0. So even after normalization, our votes continue to be exact antivotes of each other.

The price of lack of precinct sumability is worth paying to get the correct candidate in office. Bad governance is very expensive, possibly causing early human extinction, a very high price in my account book. In a given round, the precincts can report in their totals. The center then determines whom to eliminate and broadcasts this back to the precincts, which can then perform normalization and continue into the subsequent round.

Normalizing Elimination Voting <-- possible title

On Sunday, March 11, 2018 at 10:54:02 PM UTC-4, NoIRV wrote https://groups.google.com/forum/?fromgroups#!topic/electionscience/EsORYiXqhk8

[parker friedland]

This is Brian Olson's instant runoff normalized range voting: https://groups.google.com/forum/m/#!searchin/electionscience/Brian$20olson/electionscience/zu_ZMx4CWpU

[Warren D Smith]

your proposal here is similar to Brian Olson's IRNR system.

It will violate practically every "property" like monotonicity etc
people say they want. As does IRNR.

However, strangely enough, IRNR works well in bayesian regret testing.
The reason is Olson designed it in constant
feedback from a computer BR tester.



--
Warren D. Smith
http://RangeVoting.org <-- add your endorsement (by clicking
"endorse" as 1st step)

[William Waugh]

The link doesn't work.

[William Waugh]

Here's another related system.

The voters can list on their ballots so many ranks as they choose. The order of the ranks on a ballot matters.

Each rank consists of two parts: a condition and a consequence.

The condition is, or implies, a predicate over the question of which candidates are still in the running.

The consequent is a Range ballot.

A round of tallying counts, from each ballot, the consequent of the one first rank whose condition evaluates to true given the set of candidates still in the running. The candidate having the lowest total score is eliminated. The last candidate standing wins.

On Sunday, March 11, 2018 at 10:54:02 PM UTC-4, NoIRV started the conversation that you can read at https://groups.google.com/forum/?fromgroups#!topic/electionscience/EsORYiXqhk8

[Warren D Smith]

That idea by W.Waugh about "predicate choice" ballots is very interesting
from a mathematical standpoint because it is a far more general kind
of ballot expressing potentially far more information.

From a practical standpoint it sounds horrible.

[William Waugh]

A way to increase practicality might be to restrict the expression of the predicate. For example, it could be limited to one alternation or conjunction over statements that either test for a named candidate still being in the running or the named candidate having been eliminated. So, for example, you could express (Nader still in the running OR Gore eliminated), but you could not have a deeper tree like (Nader still in the running OR (Bush still in the running AND Gore still in the running)).

It would also be possible to restrict the consequent. For example, it could be an Approval ballot.

A single notation could be called for that in a combined way implies both the condition and the consequent. For example, a rank could be restricted to either support some named candidates or oppose some named candidates, and the candidates named would be the ones mentioned in the condition.

[William Waugh]

Another way to simplify the input ballots that would be evaluated as predicate ballots would be to make the voter choose an overall style from the outset. For example the choice could be IRV-style ballot or Approval ballot. An upside-down IRV-like choice could also be offered for those who want to concentrate on eliminating bad candidates rather than supporting their favorite first.

On Monday, March 12, 2018 at 12:31:29 PM UTC-4, William Waugh wrote https://groups.google.com/d/msg/electionscience/EsORYiXqhk8/7Xy79bOKAgAJ

[William Waugh]

OK, I found https://groups.google.com/forum/?fromgroups=#!searchin/electionscience/Brian$20Olson$20instant$20runoff$20normalized$20range$20voting%7Csort:date/electionscience/A__xsvjPXA8/BYlIF2RVZhEJ

and

http://bolson.org/voting/methods.html

The description is just the same.

The Yee diagrams it gives rise to look like those for IRV. So, that's not encouraging for wanting to endorse the system.

[Warren D Smith]

Olson's IRNR was not quite the same as yours,
e.g. it used L2 norms rather than your normalization notion.

[William Waugh]

I tried looking up "L2 norm" and I don't see how what comes up applies. In in simple terms about operation, how would these systems differ? The normalization meant by the OP, is that linear? Does his or her system exhibit nonmonotonicity? What other "desirable properties" does it break? Are its Yee diagrams probably forky like that for IRV?