RCVis explanations are incorrect when Bottoms Up tabulation is used for multi-winner

[davidlbangs]

Washington DC will be using Bottoms Up tabulation, meaning that candidates are eliminated until only the number of winners are left. There is no concept of a threshold. So, RCV123 happened to set the threshold to 0. Any value would be incorrect.

What support does RCVis have to tabulation methods other than STV?

Round 10 Explanation
What is happening?

These are results from a ranked choice election with 2 winners. Voters were asked to rank the candidates from most to least favorite. The ballots are then counted in rounds. In each round, the candidate with the fewest votes is eliminated, and their supporters' votes transfer to the next choice on their ballots. The rounds continue until 2 candidates have been elected.

Why were Candidate 2 and Candidate 3 elected this round?

In a ranked choice election with 2 winners, each candidate needs to earn more than 0.0% of the votes to win a seat. 0.0% is the threshold for victory because only 2 candidates can earn more than 0.0% of the votes.

Why did they need 0 votes to win?

In a single-winner election, the winning candidate needs to earn the support of at least 1 in 2 voters. Since each voter only gets one vote, the same requirement is not possible for this Single Transferable Vote election. Instead, each winning candidate must earn the support of 1 in 3 voters.

What are Inactive Ballots?

A ballot is "inactive" if the voter did not rank any of the candidates remaining in that round. Voters are not required to rank all the candidates, so if all the candidates they ranked are eliminated, their ballot becomes inactive.

[artoonie]

This is an interesting case because the RCTab JSON File doesn't contain any information about whether bottoms-up was used or not. The solution here would be to modify the checkbox for "Is this election a Preferential Block Voting election?" to instead ask the user to specify any special rules, including Preferential Block Voting and Bottoms-Up.

1. Modify the boolean to a dropdown selection box (including a migration that preserves data)
2. Modify the FAQs to explain bottoms-up
3. Add tests to ensure the FAQs are correct

I don't have the resources to take this on since bottoms-up is rare, but if DC wishes to use RCVis for official election night reporting, I'm happy to have a conversation with them.

[davidlbangs]

In the case of RCV123, you could notice that the threshold is set to zero. If so, there is no threshold.
RCTab specifies a non-zero threshold , I don't know why, and RCTab doesn't discuss that threshhold, I don't know why.

We cannot modify the checkbox , only the file, since RCVis is in an iframe.

I think it would be permissible to allow additional fields in the JSON file. Tell RCVis about the new fields, and they could choose to emit them or not.

I think it would be a good idea to allow some settings to be exposed in future RCTab report JSON files.

[artoonie]

Every option in the Upload page is also available via the API that RCV123 uses.

It's possible to discuss with the RCTab team whether the JSON should include the configuration options, but we'd have to figure out which options are important enough to include. RCVis is committed to using the RCTab JSON file as its backbone for now, since that's the format that will be used by election administrators. If we adopt our own separate format, we'd be creating features available to RCV123 that is not available to election administrators.

[davidlbangs]

New options would be adopted by RCTab if they are necessary to produce a good results screen. If you can produce proper results from all RCTab tabulation methods, then the file format would not need to be modified - RCV123 would just write the file the same way RCTab does. However, RCVis might not be able to properly describe results without knowing the tabulation method.

[davidlbangs]

Specifically, the bottoms up tabulation method in which there is no threshold , but candidates are eliminated until only the number of winners is left. How will RCVis describe the Washington DC primary results without knowing that?

Note the only reason RCV123 is adding bottoms up support it can be a proper training tool in the DC primaries - and you have to support that election anyway.

[davidlbangs]

you could modify the check box if that is sufficient, and let us know how our API call would change to designate the proper option.

[artoonie]

I haven't spoken to DC about using RCVis, I'm not sure they'll be using it, so we may not have official support for Bottoms-Up without a contract in place with DC.

In the meantime, does simply removing the "threshold" entirely lead to RCVis no longer presenting incorrect information?

[davidlbangs]

Done. threshold removed entirely for bottoms-up tabulation. RCVis stopped talking about thresholds.