If someone is favoured by a large enough group, they're supposed to win.
The point is what happens when people aren't fanatical but are forced to vote as if they were.
In a single winner election AV makes coordination among moderates somewhat easier, so it's an improvement, but still inferior to Ranked Choice/STV with multi-member districts.
https://rpubs.com/Jameson-Quinn/VSE5key
Your scenario is unrealistic because real humans aren't that tightly clustered, so you get defectors. And people vote strategically. The winning position is the centroid.
https://www.rangevoting.org/AppCW
It is not known whether proportional representation can compete with the best single winner voting methods. We don't have enough data and we can't do multi-winter Bayesian regret calculations confidently. Here's an article you really should read if you think proportion representation is the answer.
A. The research compares proportional (PR) methods to plurality voting, the worst single-winner voting method there is. But superior methods like score voting and approval voting roughly double the accuracy (i.e. group welfare) compared to plurality voting. So no, there's no research showing the superiority of PR to these methods.
B. Even in that limited comparison, of PR to plurality voting, the data still is actually not that clear. For instance, Canada uses plurality voting rather than PR, and is rated as one of the most high functioning democracies by The Economist. Meanwhile lots of proportional democracies are far worse than plurality countries. Compare proportional Brazil (and several other Latin countries) to non-proportional UK, Canada, and USA.
I don't know the metric you are using here, but it is surely purely academic, as the rest of your comment. The idea that a single candidate can better represent a smaller (and usually gerrymandered) district than a larger number of diverse candidates in a larger district flies in the face of everything we know about real democracies and party systems.
> Your scenario is unrealistic because real humans aren't that tightly clustered, so you get defectors.
Again, theoretical claims against real world experience. I've experienced polarization in AV to the point where center becomes a wasted vote, and people are incentivized to support one of the two dominant slates.
This is simply not correct. Elections are a form of statistical sampling, and with a good single-winner voting method, all voters of every stripe affect the centroid position, "tugging" it left, right, etc. Which you can see graphically depicted here.
https://www.youtube.com/watch?v=-4FXLQoLDBA
> Again, theoretical claims against real world experience.
You're not citing any real world data that contradicts my argument.
> I've experienced polarization in AV to the point where center becomes a wasted vote, and people are incentivized to support one of the two dominant slates.
If you mean "alternative vote" (known to experts like myself as "Instant Runoff Voting"), note that I'm not advocating that. I'm advocating cardinal methods such as score voting and approval voting. Which tend to elect broadly appealing centrists.
Perhaps you were using one of these variants.
https://en.m.wikipedia.org/wiki/Multiwinner_approval_voting
You should probably specify to avoid confusion.
I was mostly thinking about single-winner elections because the big headline elections in the US are all single-winner.