> Your mathematical demonstration of this is where?
http://math.temple.edu/~wds/homepage/voFdata
> With two candidates, honest range voting gets the best possible result.
Honest range voting gets the best possible result for SOCIETY under _every_ model, no matter how many candidates you have. In fact, the more you have, the more of a benefit you get by using Range Voting, as methods like plurality and IRV crash and burn with lots of candidates.
Still, from an individual's point of view, it's always better to be strategic (duh!, by definition).
> Strategic range voting is the same as single preference.
Nope. The ideal way to vote strategically with Range Voting can be quite complex, but generally has to do with maximizing or minimizing every candidate according to a general rule that I won't go into detail about here. The first four steps are as follows:
1) Maximize your favorite candidate. (Not always your best strategy with a lot of other methods, so if Nader is your favorite, you might lie and say Bush is, but NEVER the case with Range Voting.) 2) Minimize your LEAST favorite. 3) Polarize the two perceived front-runners, maximizing the one that you like more, and minimizing the other one, if if you love them both, or hate them both. 4) Go from there, maximizing and minimizing such that the sum of the pairwise differences in utility between candidates opposing each other (where one is max and the other is min) is maximized (taking revers polarity into account).
Got all that?
> With multiple candidates, it's impossible to say what strategic range voting will do, because it depends on everybody's guesses about how everybody else will vote.
But you can simulate that, for instance simulating the effect of pre-election polls, that give voters a pretty good indication of who the front runners will be. You can also use what is considered to be an EXCEPTIONALLY good and simple strategy, where you take a good guess of about how satisfied you expect to be with the winner of the election, and then you maximize all candidates whom you perceive to be better for you than that, and minimize the rest. That's the most sure-fire strategy, because it's both very good, and very reliable, and easy to calculate. It's the strategy that the simulations use.
And note that it's a good thing that Range Voting makes strategy unintuitive and kind of difficult - that helps to diminish the negative effects of strategy.
> It isn't hard to construct examples where the winner is clearly not preferred.
If by "preferred" you mean, is not the majority or Condorcet winner, then you are quite correct. So what? The winner who produces the greatest social utility is not necessarily the Condorcet winner.
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
timeasmymeasure
Re: Disaster?
Date: 2007-01-23 04:20 am (UTC)http://math.temple.edu/~wds/homepage/voFdata
> With two candidates, honest range voting gets the best possible result.
Honest range voting gets the best possible result for SOCIETY under _every_ model, no matter how many candidates you have. In fact, the more you have, the more of a benefit you get by using Range Voting, as methods like plurality and IRV crash and burn with lots of candidates.
Still, from an individual's point of view, it's always better to be strategic (duh!, by definition).
> Strategic range voting is the same as single preference.
Nope. The ideal way to vote strategically with Range Voting can be quite complex, but generally has to do with maximizing or minimizing every candidate according to a general rule that I won't go into detail about here. The first four steps are as follows:
1) Maximize your favorite candidate. (Not always your best strategy with a lot of other methods, so if Nader is your favorite, you might lie and say Bush is, but NEVER the case with Range Voting.)
2) Minimize your LEAST favorite.
3) Polarize the two perceived front-runners, maximizing the one that you like more, and minimizing the other one, if if you love them both, or hate them both.
4) Go from there, maximizing and minimizing such that the sum of the pairwise differences in utility between candidates opposing each other (where one is max and the other is min) is maximized (taking revers polarity into account).
Got all that?
> With multiple candidates, it's impossible to say what strategic range voting will do, because it depends on everybody's guesses about how everybody else will vote.
But you can simulate that, for instance simulating the effect of pre-election polls, that give voters a pretty good indication of who the front runners will be. You can also use what is considered to be an EXCEPTIONALLY good and simple strategy, where you take a good guess of about how satisfied you expect to be with the winner of the election, and then you maximize all candidates whom you perceive to be better for you than that, and minimize the rest. That's the most sure-fire strategy, because it's both very good, and very reliable, and easy to calculate. It's the strategy that the simulations use.
And note that it's a good thing that Range Voting makes strategy unintuitive and kind of difficult - that helps to diminish the negative effects of strategy.
> It isn't hard to construct examples where the winner is clearly not preferred.
If by "preferred" you mean, is not the majority or Condorcet winner, then you are quite correct. So what? The winner who produces the greatest social utility is not necessarily the Condorcet winner.
CLAY