So, a real-world story about how Ranked Choice voting by successive elimination (as implemented in San Francisco) is weird. Many following SF politics saw the initial results for Mayor last night, when only the early vote-by-mail ballots had been counted. Breed was winning.
1
7
7
Many also saw the later updates, late last night and today, with Leno winning. As of this afternoon (Preliminary Report 5), Mark Leno is barely ahead, thanks to RCV. (The election is still wide open, since there are a ton of vote-by-mail ballots still to be counted.)
1
1
What hasn't been seen much is the difference between those: the election day ballots. If you count *only* the election day ballots, Breed also wins, with Kim second. The fact that Kim eliminated Leno allows Breed to win by beating Kim; she can't beat Leno in these ballots.
2
1
Two weird things here: First, you can add two sets of ballots that both have Breed winning, and produce a result where she doesn't win. Second (& more serious), the RCV in the election day ballots eliminates Leno early when he could have beaten Breed had he stayed another round.
2
7
One clarification on the data: these are the election-day voting and vote-by-mail totals as of Preliminary Report 5, so there are some additional vote-by-mail ballots relative to the early report on election night.

Jun 7, 2018 · 6:35 AM UTC

2
Another interesting tidbit looking at only Election Day votes: The three most likely final rounds would have been: - Breed (40949) defeats Kim (40024) - Kim (36121) defeats Leno (35400) - Leno (42882) defeats Breed (39454)
1
Because the next-to-last round is Breed (34649) - Kim (26805) - Leno (23629), the final round with that set of votes (election day only) is Breed-Kim, and Breed wins the RCV algorithm. But I think that's just by luck and not because she did better in the next-to-last round.
1
That said, I think RCV is better than first-past-the-post (plurality) or having a separate runoff election of the top two. However, I'd prefer approval voting over RCV.
2
Replying to @davidbaron
is the difference in the number of ballots in each round from people not making a 3rd, 4th, etc choice?
1
Yes, or from all of their choices being eliminated, or from marking more than one candidate on one of the later choices. Also, SF only allows three choices.
1