tag:blogger.com,1999:blog-8217294657304563413.comments2020-03-21T15:06:12.521-07:00OpaVote BlogJeffhttp://www.blogger.com/profile/11828614939910702199noreply@blogger.comBlogger92125tag:blogger.com,1999:blog-8217294657304563413.post-6048056200360981112020-03-18T13:39:44.803-07:002020-03-18T13:39:44.803-07:00With ballot completion and 5 candidates, a first r...With ballot completion and 5 candidates, a first ranked candidate gets 4 points. The 4 unranked candidates split the remaining points and each get (3+2+1)/4 points. Without ballot completion, the unranked candidates get 0 points. Create a Count to try it yourself.<br /><br />Jeffhttps://www.blogger.com/profile/11828614939910702199noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-20239875461491865032020-03-18T13:17:42.054-07:002020-03-18T13:17:42.054-07:00so with ballot completion option the first candida...so with ballot completion option the first candidate would get five points and all the others 4. and it is also possible not to have the ballot completion option so that the first candidate gets five and all the others zero.harmonikhttps://www.blogger.com/profile/16041944037595313409noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-70356362898057037922020-03-18T05:49:05.937-07:002020-03-18T05:49:05.937-07:00For our Counts, we have this as an option. See her...For our Counts, we have this as an option. See here: https://www.opavote.com/help/online-counts#c-optionsJeffhttps://www.blogger.com/profile/11828614939910702199noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-55919817001294515622020-03-18T05:04:14.304-07:002020-03-18T05:04:14.304-07:00how do you count truncated ballots? If there are f...how do you count truncated ballots? If there are five candidates and a voter only ranks one candidate do they get one point and the rest zero, or do they get five points and the rest four, or do they get five points and the rest 2.5.harmonikhttps://www.blogger.com/profile/16041944037595313409noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-79382327122346806632019-11-14T12:13:09.576-08:002019-11-14T12:13:09.576-08:00Actually, looking closely at those links you provi...Actually, looking closely at those links you provided, it looks like after a candidate is elected, surplus votes are not proportionally transferred to their next choice. So the "sequential IRV" algorithm is less fair than what I show in column L.happyjack27https://www.blogger.com/profile/06322415607461417789noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-34412444593894505622019-11-14T12:04:55.546-08:002019-11-14T12:04:55.546-08:00Before I wrote the article, I wrote a program to t...Before I wrote the article, I wrote a program to test a few combinations of rules, and foudn 42 different test cases to run it on, and compiled the results into a spreadsheet:<br /><br />https://docs.google.com/spreadsheets/d/18AN5xli49l39Fh5bRQerLPF0iT5AJpEjY6dvUzk-TUI/edit?usp=sharing<br /><br />Column M is traditional STV<br />Column N is "sequential IRV"<br />Column F is "Baas results"<br />Column E (a late addition to the spreadsheet) is a modification of "Baas results" that is a lot faster (polynomial time), so is practical for elections with very may candidates. <br /><br />happyjack27https://www.blogger.com/profile/06322415607461417789noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-54153902631595295362019-11-14T09:24:55.778-08:002019-11-14T09:24:55.778-08:00In Nov 2019, the Utah county cities of Payson and ...In Nov 2019, the Utah county cities of Payson and Vineyard did multi-winner city council elections using Sequential IRV tallies.<br /><br />It occured to me that the issue of re-introducing eliminated candidates is handled by doing these subsequent IRV tallies.<br />By doing a single winner IRV, they are using 50%+1 Droop quota each time with that quota being re-calculated based upon the "continuing" ballots.<br /><br />Would be cool to see sequential IRV results plotted alongside traditional STV and Baas results.<br /><br />http://utahcounty.gov/Dept/ClerkAud/Elections/ResultsTracking.html<br />https://rcvis.com/visualize=payson-seat-1-2019-11-12_13-46-26_summary_1json<br />https://rcvis.com/visualize=payson-seat-2-2019-11-12_13-46-26_summary_2json<br />https://rcvis.com/visualize=payson-seat-3-2019-11-12_13-46-26_summary_3json<br />https://rcvis.com/visualize=vineyard-seat-1-2019-11-12_13-49-22_summary_1json<br />https://rcvis.com/visualize=vineyard-seat-2-2019-11-12_13-49-22_summary_2json<br />CJMDriverhttps://www.blogger.com/profile/15490603723527490710noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-88964527592914551652019-10-16T02:09:22.495-07:002019-10-16T02:09:22.495-07:00Our Bylaws require NOTA as a candidate but we also...Our Bylaws require NOTA as a candidate but we also allow NOTA to be eliminated so we list NOTA just like any other candidate.Anonymoushttps://www.blogger.com/profile/03850951039322644722noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-5954145570331009852019-09-17T15:04:13.971-07:002019-09-17T15:04:13.971-07:00Ah, thank you. Ah, thank you. saideshttps://www.blogger.com/profile/05177019537536865744noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-686763101492210002019-09-17T14:01:42.407-07:002019-09-17T14:01:42.407-07:00by "adjusting the max borda value" i mea...by "adjusting the max borda value" i mean that after you drop or elect a candidate, you recompute the borda counts from scratch, since now you have a different number of candidates and/or a different number of ballots.happyjack27https://www.blogger.com/profile/06322415607461417789noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-17426644872772929652019-09-17T12:06:36.599-07:002019-09-17T12:06:36.599-07:00When you say, adjusting the max borda value, I tak...When you say, adjusting the max borda value, I take it this method depends on conventional borda count rules? That is, NOT a system like Dowdall/Nauru, which would not give the same results? (the wikipedia article on Borda claims that 30% of Nauru elections would have a different winner using more conventional borda rules; also claims that the Nauru system favors higher-ranked candidates more than conventional borda, which seems reasonable to my own untrained eye--that is, likely to be true, and also possibly a good idea. I also like Nauru because I feel like it would be simpler to describe)saideshttps://www.blogger.com/profile/05177019537536865744noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-90198391836444251712019-09-01T05:26:36.516-07:002019-09-01T05:26:36.516-07:00Yes, that would work too.Yes, that would work too.Jeffhttps://www.blogger.com/profile/11828614939910702199noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-59515101493667462202019-09-01T02:17:55.295-07:002019-09-01T02:17:55.295-07:00An additional way to treat equal rankings would be...An additional way to treat equal rankings would be to randomly assign them an unambigous preference order.Martin Wilkehttps://www.blogger.com/profile/00631561664443507247noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-13370155437323486992019-08-23T14:07:31.551-07:002019-08-23T14:07:31.551-07:00Eureka!
So I decided to try the traditional elimi...Eureka!<br /><br />So I decided to try the traditional elimination algorithm, but instead of elimination the fewest first choices each round, eliminating the candidate with the lowest Borda Count (adjusting the max borda value for elected and dropped candidates, and using the re-weighted ballots). I've added the results to my spreadsheet, column E. ( https://docs.google.com/spreadsheets/d/18AN5xli49l39Fh5bRQerLPF0iT5AJpEjY6dvUzk-TUI/edit?usp=sharing ) It is very close to FEO, and completes in polynominal time (something like O(N^2), I think.)!<br /><br />So there's the solution to the issue of computing time when there are a lot of candidates.happyjack27https://www.blogger.com/profile/06322415607461417789noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-51644300284753036562019-08-23T12:18:53.633-07:002019-08-23T12:18:53.633-07:00Here's an idea:
1) for each candidate, find o...Here's an idea: <br />1) for each candidate, find out what other candidates would have to be eliminated for that candidate to make quota, with the fewest overall rank demotions. this could be about O(N^2) per candidate, so O(N^3) total for this step.<br />2) find the candidate(s) with the lowest number of such rank demotions to make quota. O(N)<br />3) if there is only one such candidate, that's the winner. if there is a tie, take the one with the most votes in their respective scenarios.<br /><br />If I'm thinkiing that through right it would be O(N^3) time, which is much better than O(N!), and would have similiar results to FEO. Someone would have to code it and test on the 43 scenarios i collected.happyjack27https://www.blogger.com/profile/06322415607461417789noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-11078854018001561262019-08-23T06:23:07.447-07:002019-08-23T06:23:07.447-07:00I'm curious what you mean by "my local co...I'm curious what you mean by "my local community". Elections for a municipality? The ballots can be entered into a computer by hand and then computer tallied. This would be more accurate and reliable than a hand tally. If people are concerned about transparency, one could provide the source code and the data (anonymized, of course), so that any one can reproduce the tally. I'm sure there are ways to have a faster elimination ordering that comes close to the accuracy of the FEO. I'm not sure your suggestion is the best approach, though. I'll have to give it some thought.<br /><br />One could use the traditional STV elimination (eliminate lowest total until someone meets quota), and then look at each dropped candidate and see if they would have won if they weren't dropped but all the others that were dropped were still dropped. e.g. if lowercase means dropped, when someone meets quota at A B C d e f, try A B C D e f, A B C d E f, and A B C d e F. then if D, E, or F win, that's the real winner. So it's basic STV elimination, with an added rollback test on each dropped candidate.happyjack27https://www.blogger.com/profile/06322415607461417789noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-86870595712156454612019-08-22T12:39:24.076-07:002019-08-22T12:39:24.076-07:00Mr. Baas, I've been very impressed with this r...Mr. Baas, I've been very impressed with this revision of the STV counting process and I'm hoping to get the shortcut version known and implemented in my local community. I don't see the full elimination ordering being sufficiently practical, unfortunately, especially if ballots are counted by hand. But I wonder if one could get any closer to the FEO result by eliminating the candidate who appears, at all rankings, on the fewest ballots, instead of the one with the fewest votes. saideshttps://www.blogger.com/profile/05177019537536865744noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-77977267700754942372019-08-12T20:05:06.136-07:002019-08-12T20:05:06.136-07:00thanks! very helpfulthanks! very helpfulSeanhttps://www.blogger.com/profile/18041290992065998206noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-29412705758702052522019-07-21T06:23:53.902-07:002019-07-21T06:23:53.902-07:00In the system I invented, which uses four averages...In the system I invented, which uses four averages or representative counts, the first average is of the Hare quota and the Droop quota, which I worked-out to be a Harmonic Mean quota (because these quotas are harmonic series): HMQ = votes/(seats + 1/2). <br />The FAB STV system does not eliminate candidates, permanently or temporarily, before the full schedule of the count is complete.<br />FAB STV extends the Meek STV use of keep values. But discards its quota reduction (also employed here). The FAB STV way is that all preferences are counted (maximum use of preference information) including abstentions, which makes quota reduction irrelevant. DemocracySciencehttps://www.blogger.com/profile/07658657506368612732noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-84417006235580935112019-03-23T11:19:59.765-07:002019-03-23T11:19:59.765-07:00You can find all Scottish Council elections for 20...You can find all Scottish Council elections for 2012 and 2017 done with both official WIG counting and with Meek STV, and switch between the two systems, at www.macs.hw.ac.uk/~denis/stv/index.html<br />I haven't indicated there which outcomes would have been different under Meek STV, I think around 1%. DenisUnknownhttps://www.blogger.com/profile/16766036513739455396noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-14070894102308805622018-12-23T20:43:24.664-08:002018-12-23T20:43:24.664-08:00That's a tough one, but I think the Borda coun...That's a tough one, but I think the Borda count is your best option.Jeffhttps://www.blogger.com/profile/11828614939910702199noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-1959746490562198002018-12-23T20:36:08.087-08:002018-12-23T20:36:08.087-08:00How appropriate is this to VERY small elections, l...How appropriate is this to VERY small elections, like for example a board of 7 electing their officers?Anonymoushttps://www.blogger.com/profile/13968435356287053377noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-67474374846691244852018-12-19T13:07:43.337-08:002018-12-19T13:07:43.337-08:00The "full elimination ordering" has some...The "full elimination ordering" has something like 5 levels of tie-breaking. I ran a bunch of simulations to optimize the tie-breaking order. (Though really the first tie-breaking criteria is exponentially more important than the others.) The best first criteria I found was "fewest rank demotions". But yeah, failing an efficient way to do that, or some other deterministic criteria, random tie breaking would be the next best. Random tie breaking can be implemented without adding a checkbox, simply by randomly sorting the candidates _before_ tallying and determining winners. i.e. randomly assigning each candidate to "A,B,C,...", instead of a deterministic method such as alphabetical.happyjack27https://www.blogger.com/profile/06322415607461417789noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-16752696251858639212018-12-18T20:03:25.470-08:002018-12-18T20:03:25.470-08:00Speaking of ties, I tried a sample election and no...Speaking of ties, I tried a sample election and noticed that your algorithm eliminates tied candidates in the order they are listed, leaving the last of the tied candidates to be winners. I suppose that there is not much that can be done with ties, but it would seem that it would be better to eliminate randomly than to eliminate in ballot order (although your method does provide better information as to what is happening). Maybe a checkbox that says "eliminate ties randomly" would be helpful. Anonymoushttps://www.blogger.com/profile/08560473952197625800noreply@blogger.comtag:blogger.com,1999:blog-8217294657304563413.post-52438192711469452362018-12-18T11:20:43.027-08:002018-12-18T11:20:43.027-08:00This comment has been removed by the author.Anonymoushttps://www.blogger.com/profile/08560473952197625800noreply@blogger.com