Lots of assumtions I didn't quite grasp. What is the distribution of voters? Some uniform space? The voters are rarely distributed that way. So I don't know what I learn from those graphics.
Each point in the image corresponds to an election with the center of opinion located at that point. For every point, we simulate an entire election by scattering 200000 voters in a normal distribution around that point and collecting ballots from all of the voters; then we colour the point to indicate the winner.
So, still confused. How is the voter ballot simulated? From a point on a normal distribution? How does that translate to their vote? Especially with some of the esoteric schemes - is their any attempt to distribute the voters' comprehension of the candidate position? Or is that folded into the voter position in the distribution? What is the width of the distribution?
Each point in the 2D space represents a person's position on two issues, one for each axis. The only independent inputs are the candidates' positions. Different voting methods have different ballots, but the idea is pretty much the same: the voters rank the candidates by proximity. So the voter would say "My first choice is the candidate closest to me, my second favorite is the next-closest, and the last one is my least favorite." So the various algorithms take those votes and choose a winner for the election.
It's described under the 'simulation method' heading. Unless you have a reason to suggest otherwise, it's reasonable to assume voters' preferences follow a normal distribution.
You are implying that there are no moderate (or 'swing') voters, which is foolish when the big parties actively reach out for these people.
Instead, issues get polarised. Both parties and candidates have several different issues, so voters' actual preferences depend on which issues are important to them.
You also get tribal voters ("My family has always voted conservative") who are unlikely to care about the issues in any election, but these cancel each other out somewhat in a plurality or AV/IRV vote.
Nevertheless, I hope we can agree that accurately modelling voting habits is complicated, not simple.
Some high-quality discussion of this from Tim Gowers (mathematician, Fields Medal winner, very smart chap): http://gowers.wordpress.com/2011/04/20/is-av-better-than-fpt... http://gowers.wordpress.com/2011/04/29/av-vs-fptp-a-suppleme... http://gowers.wordpress.com/2011/04/30/av-vs-fptp-the-shorte... (if you're only going to read one -- and I wouldn't blame you, because they're very long -- read the last one).
Some nice graphics illustrating some pathologies with AV: http://zesty.ca/voting/sim/ (note: http://zesty.ca/ is full of interesting things).
Anti-AV material by an advocate of range voting (note: he dislikes plurality voting, aka first-past-the-post, even more): http://rangevoting.org/IrvPathologySurvey.html