The RCV elimination order looks pretty stable... the closest elimination %-wise is the very last one. (If there were other close ones, that could throw things off in interesting ways... another downside of this counting method!)

Sadly we're nowhere near close enough to the true chaos scenario: Kim surging to second and getting to see what all the Leno voters put after him.

I think you can pull that out of the raw data. Maybe I'll look tomorrow.

I don't think they publish data for eliminations that didn't happen.

But they publish all the ballots! At this point I have JS code that runs the ranked choice and reproduces (on spot check of numbers) the official table. Still need to: * make it produce readable output rather than a JS object * make it show more data than the official results

I tossed what I have up at github.com/dbaron/sf-electio… ... but I have some plans for what to do next. (Right now the output is literally to console.log() because I haven't written anything else.)

Where do they publish all the ballots?

At sfelections.sfgov.org/june-5… . See "Ranked-Choice Voting: Ballot Image" and "(?)" and "Ranked-Choice Voting: Master Lookup".

So every line in "Ballot Image" is an encoded ballot?

Every line is a single rank from a ballot. Notice columns 34-36 go 001, 002, 003, 001, 002, .... (See my JS code, too.)