Instructions were unclear

[mouseas]

The program wants each ballot to rank the candidates 1-n, with each position left-to-right corresponding to a candidate top-to-bottom, but this is not clear. When I first used the program, I thought it wanted most-desired to least-desired, left to right, with the number corresponding to the candidate, rather than the position. In this way a partial ballot would simply leave off candidates from the end of their list.

Without a GUI, the instructions need to be very clear.

[PeterTheOne]

Thanks for your feedback, there hasn't been a lot of user testing. I totally agree that it may be confusing. Do you have concrete suggestions how the instructions should be altered? I made some changes already.

[Ads20000]

Do you have concrete suggestions how the instructions should be altered?

Not sure, I'll try and get round to doing some proper thinking on it :)

[cmcnulty]

I found it so confusing and difficult to use that I had to fork it - my fork allows you to enter the list indexed by order of preference - e.g. if the icecream list is: Chocolate, Vanilla, Strawberry, and your favorites are Vanilla, Strawberry, Chocolate - you enter "2,3,1" rather than what this system would call for, which is: "3,1,2". The big advantage of my fork, though, is that it allows entry of just "3" if your favorite is just Strawberry, and it will assign zeros to anything following it - which would be equivalent to an entry of "0,0,1".

[PeterTheOne]

@cmcnulty I try to explain the logic behind the current way of doing it. This IRV site was created to help count paper ballots and thus optimizes for transcribing paper ballots. A paper ballot has a list of Chocolate, Vanilla, Strawberry with "check boxes" next to them. Reordering on paper doesn't work, so you write the order of candidates in the boxes. When transcribing you should not think about the ballot anymore but just transcribe what you see, therefore you just transcribe the numbers from top to bottom on the ballot.

Your input format is certainly more intuitive to understand, I think it could be even better when you replace numbers with the actual candidate names! But unfortunately it doesn't line up with what I wrote before. If you like, we could merge something like that as an alternative input method.

[cmcnulty]

I completely understand now - it's all about how the ballots are distributed and how the votes are collected - I optimized for the case where, for instance, a numbered list is sent out via e-mail, with instructions to send back a list of numbers in the order of preference.

Your code is optimized for a situation with a printed ballot with instructions to add a number next to each line on the ballot indicating order of preference.

In the end the code to translate between the two methods is pretty trivial - I think the best direction for this project, to be honest, would be to split the vote handling code into it's own nodeJS module, and handle the ballot parsing in a separate UI only project.

[hfossli-agens]

Candidates

Obama
Hollande
Merkel

Ballots

3, 1, 2
3, 1, 2
1, 2, 3
1, 2, 3
2, 3, 1
2, 3, 1

Actual result

Round #1:

3 candidates and 6 ballots.

Number of first votes per candidate:
Obama: 2
Hollande: 2
Merkel: 2

3 candidates have the highest number of votes with 2 votes (33.33%)
3 candidates have the lowest number of votes with 2 votes (33.33%)

Tiebreaker: Hollande was randomly selected as the loser of the round.

Round #2:

2 candidates and 6 ballots.

Number of first votes per candidate:
Obama: 2
Merkel: 4

Merkel has the highest number of votes with 4 votes (66.67%)
Obama has the lowest number of votes with 2 votes (33.33%)

Merkel won!

Expected result
I didn't expect Obama and Merkel to have any difference in round 2. They seem to have same number of no 1 votes and same number of no 2 votes.

If this issue can be dismissed by my lack of knowledge - feel free. To me this seems wrong.

[PeterTheOne]

@hfossli-agens As Hollande lost round 1 now for every ballot with Hollande as "first votes" the "second votes" will become the new "first votes". In this case Merkel gets two extract "first votes". Did you expect it to remain a tie? It is the concept of IRV to still consider the preferences of ballots that didn't get their first wish.

[hfossli-agens]

I expected obama and merkel to tie with 4 - 4 in round 2 and a new loser be selected randomly. I might be uninformed. Håvard

…

27. nov. 2020 kl. 18:42 skrev Peter Grassberger ***@***.***>:  @hfossli-agens As Hollande lost round 1 now for every ballot with Hollande as "first votes" the "second votes" will become the new "first votes". In this case Merkel gets two extract "first votes". Did you expect it to remain a tie? It is the concept of IRV to still consider the preferences of ballots that didn't get their first wish. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.