I’m glad you enjoyed the post.

I didn’t include analysis of STAR because it failed IIA, however I can understand why some people don’t consider this a deal-breaker – the fact that tactical voting can make cardinal methods fail IIA does reduce the impact of the failure somewhat.

Perhaps I wrote it off too quickly and didn’t do it justice.

There are a few more disadvantages to STAR though:
1. Although the monotonicity criterion is not failed by STAR in a technical sense, the situation hhh links to above is nevertheless quite counter-intuitive, and looks quite like non-monotonicity
2. It fails the Participation criterion, so people could be better off if they didn’t vote (as does Majority Judgment)
3. It fails the Consistency criterion, so if the electorate is split into parts, even if all of the parts have the same winner, a different winner can be elected by the combined electorate, which could lead to some very confusing maps being produced that shake people’s faith in the legitimacy of the method (again, Majority Judgment has this issue too)
4. It is a significantly more complicated method than Approval or Score voting for both voters and vote counting

Fundamentally, the “automatic run-off” seems to introduce the possibility for quite a bit of strange behaviour.

Your comment states that STAR performs better when people vote tactically, but hhh’s link suggests that >75% of voters must be voting tactically for this to be the case. In the presence of such levels of tactical voting, its tactic-resistance may well outweigh the disadvantages above, but such high levels of tactical voting sound quite unlikely to me.

I would be interested in any evidence or studies showing that either:
a. STAR outperforms Approval or Score voting with levels of tactical voting <75%
b. The proportion of people that would vote tactically under Approval or Score voting is greater than this threshold

It is possible that I could get behind the idea of score voting being used for a “primary”, with a separate vote for the final 2 run-off (in a similar way to the French presidential elections 2-round system). Making the run-off election a completely separate vote appears to avoid a lot of the strange behaviour of STAR whilst maintaining the tactic-resistance. This would be more expensive and complex, but that could be justifiable for something like a presidential election.

You are clearly someone that has looked into this in great detail, and some evidence/study/simulation/experience has convinced you that the benefits of STAR outweigh its drawbacks. I would be very interested to know what that is, as to me the combination of complexity, failing IIA, failing participation and failing consistency give it a very high barrier to overcome.

I have now edited the post to include a couple of paragraphs on STAR Voting, using some of the points above.