I just read a good journal entry about "Why Third Party Candidates Can't Get Elected" by tsg ... However one thing I disagree with is using Instant Runoff instead of one of the Condorcet voting methods.

Even though Arrow's Impossibility Theorem proves the impossibility of designing rules for social decision making that obey a number of 'reasonable' criteria that doesn't mean we can't evaluate voting methods according to various criteria.

When we do that we find that instant runoff has serious problems such that it is arguably no better than the USA's current plurality system at expanding the two-party system and giving other parties a chance to actually win elections.

Condorcet voting using the Cloneproof Schwartz Sequential Dropping (CSSD) method to resolve cyclic ambiguity is about the best all around method people have been able to come up with. Some weaknesses of the Condorcet method are that it doesn't always follow the Participation Criterion and it is still susceptible to some Strategic Nomination concerns. However in most situations it shouldn't cause a problem, and would be an incredible improvement over what we currently have.

Of course a good debate about Approval voting... maybe use one for the primary and another for the real election? They might be able to complement each other well.