User:Randomstaplers/Article 1: Difference between revisions
Content deleted Content added
copied from reddit |
update lead section... and more i guess |
||
| Line 1: | Line 1: | ||
:: ''Originally titled'' '''''[https://old.reddit.com/r/EndFPTP/comments/lq6tk3/lets_simulate_irv/ Let's Simulate IRV!]''''' |
|||
::: ''Everything's broken. I'll fix it later; more revisions to come!'' |
|||
| ⚫ | I've got a feeling that those pushing IRV (also known as RCV, or the CGP Grey 'alternative vote') at any cost, 'cause we need voting reform now!' have never actually seen IRV simulated, outside of charts provided by Grey. Because if one were to see visual aid on who wins and loses, one might be reluctant to adopt IRV. |
||
:: ''Originally called'' '''''Let's Simulate IRV!''''' |
|||
| ⚫ | |||
== Votesim == |
== Votesim == |
||
:: ''See also: [[User:Randomstaplers/Running_votesim-based_programs|Running IEVS on your computer!]]'' |
|||
The simplest computer simulation |
The simplest computer simulation, as far as I know, only requires grade school geometry to understand. It was invented by Ka-Ping Yee in 2005, no relation to any voter advocacy group... so far. You can see his results here.<ref>Explanation from Votesim's creator + images: http://zesty.ca/voting/sim/</ref> Votesim was soon added to Warren Smith's IEVS. |
||
| ⚫ | |||
The easiest way to generate your own images is to use Warren Smith's IEVS, which contains Votesim. 92k download here. You'll need to use the command line. If you're compiling, you'll need to #define HUGE and the source code has broken RNG, but it is more up-to-date, contains score-runoff voting and strategic voting settings. |
|||
== A few haphazard votesim results == |
|||
| ⚫ | |||
==== But first: A word on models ==== |
|||
:: ''See also: [[User:Randomstaplers/An_Overview_of_Voter_UI|Psychology and Voter UI]]'' |
|||
All models break down in some way, but that doesn't mean their results aren't useful. The world, for instance, doesn't look anything like what is shown on mapping websites, but it doesn't stop us from using them to navigate the world. This model assumes everyone is honest, so Borda looks a lot like score, because it cannot differentiate the difference between ranked ballots and cardinal ballots, a UI problem which can surface when real people vote. |
|||
It is also biased towards Llull winners (which Condorcet rediscovered in the 1700s, who knew?) |
|||
== More Unrealistic Simulation Examples! == |
|||
The threshold for approval voting does not correspond with polling.<ref>Approval voting scales: https://www.counted.vote/diving_into_our_straw_poll_data_subjective_scales/</ref> |
|||
All models break down in some way. (See: Google Maps.) This model assumes everyone is honest, so Borda looks a lot like score, even though it has issues, (See point 2) because it cannot differentiate the difference between ranked ballots and cardinal ballots. It is also biased towards Llull winners (which Condorcet rediscovered in the 1700s, who knew?) |
|||
==== Results? ==== |
|||
Edit: 9_Point_Buck points out that polling shows the approval threshold is different compared to what Votesim uses. Another reason models are broken. |
|||
To be uploaded... |
|||
All of the following methods have standard deviations of 50 and contain 16 candidates. Top-2 and IRV have another image with a deviation of 100 and 5 candidates. Another reason this model breaks. |
|||
| ⚫ | |||
One voter. One voter 2. The ideal model for this sim. And if you use a Llull method, you'll get the same result, even when hashed. |
|||
| ⚫ | |||
FPTP. |
|||
| ⚫ | * There are situations where IRV produces worse results than Top-2.<ref>T2R vs IRV: https://rangevoting.org/TTRexec.html</ref> And in the multi-candidate simulation, IRV doesn't produce significantly better results than Top-2. What's the point in advocating for IRV in a Top-2 state, when there are better voting systems? |
||
| ⚫ | |||
Approval. It did pretty well with 5 candidates, but 16 is just too much :( |
|||
Let's say [Red], [Yellow], and [Green] are running in a ranked election.<br /><br />[Red] does well, but doesn't reach a majority, and causes [Yellow] to lose...<br /><br />[Yellow] is evenly split. (sic, but you can't determine how the vote will go) |
|||
Range/Score |
|||
| ⚫ | |||
Top 2 FPTP and Top 2, 2 and IRV and IRV 2. |
|||
| ⚫ | |||
| ⚫ | |||
| ⚫ | |||
| ⚫ | |||
| ⚫ | |||
| ⚫ | |||
Let's say [Red], [Yellow], and [Green] are running in a ranked election. |
|||
[Red] does well, but doesn't reach a majority, and causes [Yellow] to lose... |
|||
[Yellow] is evenly split. (sic, but you can't determine how the vote will go) |
|||
| ⚫ | |||
| ⚫ | |||
| ⚫ | |||
== At the end of the day... == |
== At the end of the day... == |
||
| Line 53: | Line 42: | ||
I hope we take a more cautious approach to voting reform and not just jump on the IRV bandwagon. |
I hope we take a more cautious approach to voting reform and not just jump on the IRV bandwagon. |
||
There are better voting methods, some of which don't produce '''worse results than T2R.''' Let's not shoot ourselves in the foot here, advocating for ''any'' voting system that could do worse than '''doing FPTP twice.''' |
|||
== TL;DR for commenters: == |
|||
If you're concerned about model accuracy complaints from others, just point out Google Maps as an inaccurate model. Doesn't mean it's not useful. And yes, there are more models out there, they just require higher level maths... |
|||
Yes, this is targeted towards reformers (and the laymen) who haven't really thought of what they're promoting. |
|||
== See Also == |
|||
IMO, IRV's biggest problem is opinion point 1, that it may be worse than Top 2. If there was no runoff, I'd be in favor of it (reluctantly). But I live on the West Coast, which has a Top 2 primary. I'd probably vote against it. Link, and additional discussion. |
|||
* [[User:Randomstaplers/Running_votesim-based_programs|Running votesim-based programs]] |
|||
* [[User:Randomstaplers/An_Overview_of_Voter_UI|Psychology and Voter UI]] |
|||
== Links == |
|||
Edits: Added the approval disclaimer, Google Maps example, and the section above. And more corrections. I'm going to license this post CC0 public domain, so feel free to change things and not cite me. And the bitmaps too. |
|||
{{reflist}} |
|||
Revision as of 21:13, 12 March 2021
- Originally titled Let's Simulate IRV!
I've got a feeling that those pushing IRV (also known as RCV, or the CGP Grey 'alternative vote') at any cost, 'cause we need voting reform now!' have never actually seen IRV simulated, outside of charts provided by Grey. Because if one were to see visual aid on who wins and loses, one might be reluctant to adopt IRV.
Votesim
- See also: Running IEVS on your computer!
The simplest computer simulation, as far as I know, only requires grade school geometry to understand. It was invented by Ka-Ping Yee in 2005, no relation to any voter advocacy group... so far. You can see his results here.[1] Votesim was soon added to Warren Smith's IEVS.
How does it work? Well, Mark Frohnmayer rewrote Votesim again, this time by adding animations. His video[2] pretty clearly describes Votesim's operation.
A few haphazard votesim results
But first: A word on models
- See also: Psychology and Voter UI
All models break down in some way, but that doesn't mean their results aren't useful. The world, for instance, doesn't look anything like what is shown on mapping websites, but it doesn't stop us from using them to navigate the world. This model assumes everyone is honest, so Borda looks a lot like score, because it cannot differentiate the difference between ranked ballots and cardinal ballots, a UI problem which can surface when real people vote.
It is also biased towards Llull winners (which Condorcet rediscovered in the 1700s, who knew?)
The threshold for approval voting does not correspond with polling.[3]
Results?
To be uploaded...
So why is IRV controversial?
Opinions ahead!
- There are situations where IRV produces worse results than Top-2.[4] And in the multi-candidate simulation, IRV doesn't produce significantly better results than Top-2. What's the point in advocating for IRV in a Top-2 state, when there are better voting systems?
- IRV is absolutist. You support candidate 1, and only 1, until their elimination. Then support for candidate 2, etc. If a compromise candidate is eliminated, well...[5]
Let's say [Red], [Yellow], and [Green] are running in a ranked election.
[Red] does well, but doesn't reach a majority, and causes [Yellow] to lose...
[Yellow] is evenly split. (sic, but you can't determine how the vote will go)
So Green wins instead. Even though polling shows you would probably prefer Yellow,[6] as the optimum candidate for society.
- Laboriously transferring votes when a candidate is eliminated? No easy way to count in precincts? Less transparency is the last thing we need when people don't even trust the simple FPTP system.
- IRV has been repealed before. The fact some of us are still advocating for it is insanity. (Insert your own word here, I got tired...) Oh, and if the big parties catch on to IRV's flaws, they might use the voting method as an excuse to never reform voting again!
At the end of the day...
I hope we take a more cautious approach to voting reform and not just jump on the IRV bandwagon.
There are better voting methods, some of which don't produce worse results than T2R. Let's not shoot ourselves in the foot here, advocating for any voting system that could do worse than doing FPTP twice.
See Also
Links
- ↑ Explanation from Votesim's creator + images: http://zesty.ca/voting/sim/
- ↑ Equal Vote's animated votesim: https://youtu.be/-4FXLQoLDBA?t=20
- ↑ Approval voting scales: https://www.counted.vote/diving_into_our_straw_poll_data_subjective_scales/
- ↑ T2R vs IRV: https://rangevoting.org/TTRexec.html
- ↑ Source for the quote: https://youtu.be/vppgodFbZ84?t=324
- ↑ Yellow as consensus: https://psephomancy.medium.com/a-majority-of-voters-1d990a53b089