Algorithmic Asset Voting, a Condorcet method that limits Favorite Betrayal incentive

Continuing the discussion from:

AAV (Algorithmic Asset Voting) is just a Condorcet method where we simulate negotiators trying to strategically maximize voter satisfaction with a Smith Set candidate, one of whom must win. We run another Condorcet method to determine a “default winner”, then have the negotiators attempt to elect someone else by flipping votes around.

The standard Warren Smith examples of Favorite Betrayal failure in Condorcet are solved by this method, since the negotiators can just flip the voters’ preferences in such a way as to ensure they don’t have to Favorite Betray i.e. the method simulates FB on their behalf. The only thing required to do this is knowledge of voter preferences within the Smith Set, and since we generally expect the Smith Set to only go up to 3 members most of the time, this doesn’t seem far from precinct-summable.

I’d invite some kind of counterexample showing FB incentive with this; if you want clarification on how the method works, just ask. I’d suggest using Schulze to determine the default winner.
Possibly an equally valid way of doing this is to skip the default winner concept and give each member of the Smith Set an equal probability of winning (randomly pick if no Condorcet winner can emerge from negotiations), since that seems most analogous to how an Asset negotiation might work.

I suppose I should also mention in this post that AAV can be treated as a Condorcet PR method which maximizes Vote Management in favor of voters similar to Schulze STV. In the multiwinner case it should also be possible somehow to have negotiators flipping preference to change the Smith Set, though this is very computationally expensive I imagine.

Edit: Turns out this idea doesn’t work as is, because there’s never a stable winner, since a majority can always do Favorite Betrayal to get someone else, since it’s a Condorcet cycle.