File talk:Alternative IRV counting flowchart.png

I have nothing strong against this image change, but I wanted to explain the difference if anyone else wants to offer feedback:

The original Image:IRV_counting_flowchart.png says:

Recount all ballots

This new one says:

Distribute eliminated candidate's ballots

The reason I put "recount all ballots" in the original version was to emphasize the recount affects all ballots, while distribute more easily offers a false interpretation that some people get a second vote, while others only get one vote. "recount all ballots" to me makes it more clear that we have a "completely new election round" even if by efficiency (of hand-counting especially), you may want to "pile ballots" by vote rather than merely tallying a new count, and then you can "redistribute" as sugggested.

SockPuppetForTomruen (talk) 17:26, 16 May 2008 (UTC)[reply]

I made the change because "recount all ballots" is false. This is not what is done. --SmokeyJoe (talk) 22:36, 16 May 2008 (UTC)[reply]

How do you know that? What procedures and where? If they're done by scanners, I'd expect they'd all be run through, once for every round, counting the highest active candidate each time. If they're done in software, again, they're ALL recounted. As best I can tell, it's only in hand-counting where ballots are physically separated by votes and reditributed. SockPuppetForTomruen (talk) 03:51, 17 May 2008 (UTC)[reply]

http://www.aec.gov.au/Voting/counting/hor_count.htm, for example.

This “recount all ballots” sounds like an attempt to exaggerate the difficulties. Even if it was done electronically, it would be a crazy algorithm that would “recount” from zero. You’d probably need a special program, or basic, to prevent a compiler from optimising something so silly. In any case, where I can find good references for what *is* done, the ballots for eliminated candidates are counted separately and added to previous totals. --SmokeyJoe (talk) 00:51, 21 May 2008 (UTC)[reply]

This is silly. I've programmed IRV/STV and many EMs. There's no such thing as optimizing counting by a compiler - it depends 100% on your data structures and algorithms. If you have a single list of ballots, it must be scanned for each round of counting, simply flagging which candidates are active and counting the highest active on each ballot. If you wanted to optimize speed, you might move ballot records around between multiple lists, but a more complex algorithm, and moving ballots around between lists means more chance of programming errors and less transparancy to verify the coding. SockPuppetForTomruen (talk) 17:51, 21 May 2008 (UTC)[reply]

According to what I’ve read/is referenced, the AEC sorts ballots, then counts, and in later rounds eliminated candidates ballots are separately sorted, counted, and these totals are added to previous figures. They do not recount for each elimination, although ballots are recounted for verification. Can you provide references for the computerised IRV ballot counting that involves multiple re-scanning of ballots? Out of interest, I’m curious as to how long it takes, how often things jam, and how often ballots get ripped, crumpled or otherwise damaged by machines. Or are you talking about a hypothetical ballot count? --SmokeyJoe (talk) 00:14, 22 May 2008 (UTC)[reply]

It's largely assumption on actual procedure on my part.

On physical counting by machine, I could imagine counting machines could sort ballots by first choice, output in different stacks, which would have eliminated candidate sets be "re-feed" into the machine for redistribution. However it seems simplest mechanically to just feed in all the ballots, and get out a tally of votes for each candidate. Then flag what candidate is eliminated and refeed all the ballots for a new count. I certainly could see an advantage to actual physical grouping of ballots - so human recounting (or interpretation of uncertain/damaged ballots) would be simpler.

I suppose on a programming level I never considered redistribution because one-pass recounting per round is trivial, and I could recount with different filters, like Condorcet pairwise, or trying different elimination order permutations in the case of a last-place tie to see if it made any difference.

Well as I said from the start, my recount all ballots intention was clarity that everyone's vote counted each round. Redistribution is a convenient way of tabulating and visualizating transfer votes, but I'd call redistribution a convenience rather than a logical necessity to the method. SockPuppetForTomruen (talk) 01:25, 22 May 2008 (UTC)[reply]

A reminder, I copied the flowchart from the Minneapolis IRV campaign [1]

I made an original chart for a top-two IRV election. Image:IRV-toptwo_flowchart.png which uses less explicit language "recount votes among top-two".

Perhaps a simple Recount votes among remaining candidates is most accurate, ignoring details whether recounted batch-style, or redistributed from some sort of movement of a subset of physical ballots.

SockPuppetForTomruen (talk) 02:45, 22 May 2008 (UTC)[reply]

distribute elimated candidate's ballots

I don’t think “distribute eliminated candidate’s ballots” is accurate. Shouldn’t it say something about the second preference or the next preference?

Perhaps something like: “Distribute votes for the next most preferred candidate from the eliminated candidate’s ballots.”

I know it is more prolonged and more awkward, but isn’t it more accurate? Maybe some other shorter description? myclob (talk) 18:55, 29 December 2021 (UTC)[reply]