Loomio
Sun 5 Mar 2017 6:35PM

Score Ranked Pairs

NH Nathan Hunter Public Seen by 362

This is a repeat of the method I proposed in the initial Wonk Talk thread

SRP

Score Ranked Pairs is a Condorcet Method that uses a score ballot. Like all Condorcet methods, it will always elect the Condorcet Winner if one exists. If there is no Condorcet winner, though, it will basically defer to the highest scored candidate. Like SRV, it assumes that every precinct provides the sum of each candidates scores, and the degree of preference for every pair of candidates.

The Algorithm

Find the Top Contenders

  • Find the candidate with the highest score and add them to the set of Top Contendors (TCs)
  • Examine the preferences to see if there are any candidates that could beat a TC in a one-on-one race that are not already in the set of TCs. If so, add them to the set of TCs
  • Repeat the previous step until you find no new candidates. #### Short Circuit If there is only candidate in set of TCs, then you have your winner and you don't need to bother with the next step. #### Build the Graph For every pair of top contenders, and an edge to a directed graph pointing towards the candidate who would win in a one-on-one race. If adding such an edge created a cycle, remove the edge pointing away from the candidate in the cycle with the highest score. #### Find the Winner The winner is the root of the directed acyclic graph resulting from the previous step.

I recently found out that this equivalent to "Black's Condorcet Method", which uses a Borda Count if there is no Condorcet Winner. I like this method over traditional RP because using a score ballot could be more compactly presented while still favoring rankings over scores. I also think it's pretty safe to assume that the set of Top Contenders in an election grows more slowly than the number of overall candidates in a race, so this presents a significant gain in time efficiency, while also creating a more readable graph so that it's easier to describe the electorate's preferences.

Scope of this Thread

I remain convinced that-- at least for recurring political elections-- it is more important to elect a Condorcet Winner if one exists than to maximize Voter Satisfaction Efficiency. I know many of you disagree with me on that, and I would be willing to discuss it with you, but I think that is a topic that deserves it's own thread. There's still plenty to talk about within the framework of preferring the Condorcet Winner: how to describe the system more succinctly, it's vulnerability to strategic voting, potential refinements.

CS

Clay Shentrup Mon 6 Mar 2017 3:03AM

I remain convinced that-- at least for recurring political elections-- it is more important to elect a Condorcet Winner if one exists than to maximize Voter Satisfaction Efficiency.

This is not subjective. It is logically proven that the Condorcet winner (or even a majority winner) is not necessarily the group's favorite.

Maybe you're talking about some political practicality issue. But as far as moral/theoretical correctness goes, this is a solved problem.

WW

William WAUGH Wed 5 Apr 2017 3:40PM

If my intuition doesn't fail me in this case, every vote have a balancing vote.

Have you constructed an example election in which the result of this system differs from the result of straight Score?

(I live in Virginia)