Hacker News Reader

Comment by anonymoushn | Hacker News Reader

One desirable property of FPTP is that if you want to maximize the chances of your candidate being elected, you should vote for your candidate first before all other candidates. In IRV, sometimes you should not vote for your candidate first before all other candidates. Is that okay?

Here are some interesting diagrams of the results of applying various voting algorithms including IRV: http://zesty.ca/voting/sim/

View on HN

Those diagrams aren't very intuitive to me. Do the different shaded regions represent turnout? What is the coordinate of the candidate, position on a 2-axis political scale?

A point on the 2d plane represents a preference (if we imagine that there are only 2 real-valued issues in the world). The regions shaded in a particular color are the regions where a normal distribution of popular preferences centered there will elect the corresponding candidate.

I assume that individual simulated voters just rank candidates from least to greatest by cartesian distance, or do that and then use a random cutoff for approval voting.

There is no reason to put your candidate as lower than 1 in IRV. The more votes they get as the first preference the less likely they are to be eliminated and the more likely they are to get preferences from already eliminated candidates.

Voting for someone other than your favorite first can change the order in which candidates other than your favorite are eliminated. This can cause your favorite to win instead of losing.

http://www.rangevoting.org/ClayIrv2.html

Really cool. The link only looks at single-candidate elections, though. It's unclear to me that this is superior to party-based systems.

Particularly Congress could use some proportional representation.

EDIT: Wow, would never have guessed that IRV gave such strange results...

Any voting system that doesn't satisfy the Condorcet criterion gives what appear to be strange results.

Reply on news.ycombinator.com