> From: Andrew Plotkin <erkyrath@xxxxxxxxxx>
(If you play exactly two games, and mark A in first place and B in "tenth
place", that's just second place. Either way, it boils down to "I think A
beats B.")
The thing about Condorcet is that its good at figuring out who you want, but bad at figuring out who you really don't want - if the truth of my opinion is "I like A, sort of like B, and only kinda like C", then I would submit the ranking (A, B, C). But if the truth of my opinion is "I like A, and really, really hate both B and C", I would still submit (A, B, C). That doesn't seem quite right.
What if instead of merely ranking the games, you indicated relative amounts by which each game beats the game below it? There are a lot of ways this could be done. A simple one would be to have there be X*N slots, where N is the number of candidates, and votes are weighted not only by who they lose to, but by how much. X could be anywhere from 1 (allowing incomplete ballots to be more accurate) to 10 or so. A more fine-grained approach that would also allow ties on a single ballot would be to simply assign each candidate a number from 1-100. To determine the winner out of a given pair of candidates, instead of checking who won more times, you would check who won by a greater total amount.
This does make for somewhat counterintuitive voting - in a two-candidate election, ranking them 100 and 99 is the same as ranking them 2 and 1. But this isn't wrong. If you really hate both candidates, or if you really like both candidates, your ballot should affect things very little. If everyone understands how it works, I think this might provide a more accurate picture of what people want.
Jake
--
A warb degombs the brangy. Your gitch zanks and leils the warb.