I like these for instance:
mossy earth
Planet Wild

Ist das heute nicht auch rechtswidrig? Ich bin mir nicht sicher was genau die Regulierungen sind, aber wenn das Jungendamt heute von sowas Wind bekommt, kannst auch Probleme haben, glaub ab einem bestimmten Alter der Kinder? Was natürlich super ist, wenn du dir mehr nicht leisten kannst, und dann dafür bestraft wirst. Das wird das es-sich-leisten-können auch nicht verbessern...

amortized per diem (considering leap days) is closer to 42.71 giving

• 1195.89 / 28
• 1238.60 / 29
• 1281.31 / 30
• 1324.02 / 31

Assuming the tenant stays for more than four years

STV-CLE is still path-dependent, right?
Also, I'm specifically asking here about modifying such methods into a party form.
I think STV implementations for large parliaments generally tend to work by having smaller constituencies with a limited set of options run for a subset of the available seats to make it, like, human-tractable? Or else you only fill a subset of seats, rather than All Of Them or something like that: It's just not feasible to reasonably vote for possibly hundreds of seats, distinguishing between even more hundreds of candidates.

Like,
in the USA has 435 + 100 members in its bicameral congress, and relative to its size, that's a small number
The EU as a whole has 720 + 27
In the UK we're talking 650 + 751.
In China, democratic status not withstanding, it's unicameral with 2878...

I'm talking about an election where the entire combined constituency gets to elect the entire parliament (or at least one chamber) all at once.
It will just not be reasonably possible for every voter to have an informed

But if I'm not mistaken, having more local sub-elections implies the possibility of Gerrymandering.
I think I'd rather have a method where you select parties, but do so "all at once". In such a setup, a party method will be necessary. STV and variants is built to deal with individuals, each of who get a single seat. They aren't designed for parties where multiple seats can go to the same party...

My hope would be that electing parties in a Condorcet-PR fashion causes parties that are better aligned with the public and better able to agree when they ought to agree, as you no longer have to worry as much about things like the Spoiler Effect. This already is a little possible with STV, but there still are issues with weird election orderings, where everybody's second choice (but nobody's first) gets eliminated right away and so never has a chance of winning.

I suppose STV-CLE does solve that deficiency at least: Everybody's second choice isn't usually gonna be the Condorcet Loser.

Either way though, that variant doesn't solve the party list aspect of it.

Lied about loving all basic shapes. He is terribly shapist about sandwiches (triangles - the only shape they should be) and square pizza

utterly necessary for evolution

Markdown is cool when it works but sometimes it's such a nuisance. Sorry about some of that enumeration being broken

I kinda don't care that much about arguments of how well people understand any given system. I can guarantee you that people don't generally really understand most proportional methods already widely in use today. It's really not the kind of obstacle that it's always made out to be.

For policy makers who'd actually champion any given method, fair enough, if they are half way competent, they would have to understand it well enough to make a reasonably informed decision on this. But that's also very frequently what's not happening (with any legislation, not just with voting reform type stuff)

What this says, though is roughly this:

Say you have three ballots

1. A B C D E
   
2. D C B A E
   
3. E D C B A

and are looking to fill three seats, so we have to differentiate between one of ten possible arrangements, picking one winner:

A B C
A B D
A B E
A C D
A C E
A D E
B C D
B C E
B D E
C D E

The Droop quota in this case is floor(3/(3+1)) + 1 = 1 (very simplistic case, but this quota business is pretty common across loads of proportional representation methods that actually do get used)

We pairwise compare the above arrangements:

ABC vs ABD

1. exclude candidates in neither arrangement. E is in neither arrangement, so this gives us:

• A B C D - top choice A
  
• D C B A - top choice D
  
• D C B A - top choice D

1. transfer surpluses of candidates in both outcomes:

Between ABC and ABD, A and B exist in both.

A has exactly quota (1 vote), no surplus
B has 0 votes, no surplus

We therefore have the following counts:

A 1
D 2
B 0
C 0

1. add the totals of each out come

ABC: 1 + 0 + 0 = 1
ABD: 1 + 2 + 0 = 3

1. biggest is the winner of this comparison. ABD > ABC.

Next comparison, ABD vs ACD

E is excluded again, same candidates overall. So we once again have:

1. A B C D - top choice A
   
2. D C B A - top choice D
   
3. D C B A - top choice D

Now the surpluses:
A has 1 vote, no surplus.
D has 2 votes, the quota is 1, so we have a surplus of 2 - 1 = 1. D's surplus must be transferred.

The relevant ballots both have C as a next choice.

Since there are two ballots that generated this surplus of 1, we need to transfer 1/2 to the next choice on each ballot. Since that happens to be C both times, we get 1/2 + 1/2 = 1

As a result, after this transfer, we get:

A 1
D 1 // one surplus vote redistributed
C 1 // got two half surplus votes from D
B 0

Outcome ABD then gets 1 + 0 + 1 = 2 votes.
Outcome ACD gets 1 + 1 + 1 = 3 votes.

ACD > ABD

Thus far we have

ACD > ABD > ABC

Comparing ABC and CDE

1. Exclusions: None. Every letter exists. Ballots look like:

A B C D E - top choice is A D C B A E - top choice is D E D C B A - top choice is E

1. Surpluses: C is the only common letter but has no top votes. So no surpluses.

Outcomes look like:

A 1 B 0 C 0 D 1 E 1

1. add totals:

ABC: 1 + 0 + 0 = 1 CDE: 0 + 1 + 1 = 2

CDE wins

so thus far we have

ACD > ABD > ABC, CDE > ABC

and so on. In this extremely simple toy model, the obvious winner is ADE, where everybody gets their top choice elected. But in a more realistic setting, this obviously may not be so.

Of course, you can, and in a more realistic situation, likely will end up with Condorcet cycles. However, at that point we'd already have the rankings which can be resolved in the same ways various Condorcet method do so, using for instance Tideman Ranked Pairs or Schulze Beatpath.

Ultimately, CPO-STV is a lot of work, but it's not that complicated "per step". For each pair of candidate set and each ballot, you find what's not in either candidate, discard those options from the ballot, and then for the rest you just count up how well the current top choices match, transferring surplus votes (i.e. more votes than needed to meet the quota) to the next choice on each ballot. Once everything is transferred, you just tally up the effective vote count for each outcome, find which one wins, and construct an order, much like regular single-winner Condorcet methods would. Tie breaks similarly happen via some single-winner scheme.

The most complicated step, imo, is the quota transfer. But that's just a standard feature in loads of PR methods in general.

What about exacting and ultra exacting?

I feel like my post already answered your question, but I'll try:

Multi-winner Condorcet methods do exist in the form of CPO-STV or Schulze-STV.
Schulze-STV in particular actually reduces to D'Hondt's method if you have every voter rank only candidates of a single party, and rank them all the same, i.e. if you pretend the Ranked Choice ballot is a party-level FPTP ballot instead.
It also reduces to the (single winner) Schulze method if you are trying to fill only a single seat.

One issue with that method, as far as I can tell, is mostly the computational complexity: You compare all possible sets of seats, and that's what you apply the Condorcet method to. The goal is to find the best supported set of seats.
For any sizeable election, that is a lot of comparisons to make, so it's gonna be very slow for parliament-scale elections.

Another issue is simply what it expects of a voter:
For a single winner election with like five candidates, it's not that hard to have a reasonably informed opinion about all five candidates and to rank them accordingly.

But if you are trying to fill a 200 seat parliament, that means there are going to be at least 201 candidates to consider. Any less and there wouldn't be a need for an election at all. That would be a ridiculous ballot to fill out and almost nobody is going to go through the trouble of ranking all 201 candidates. And that's the very smallest case. More realistically, we are looking at ranking like a thousand people most of whom most voters aren't even going to have heard of.

So what we'd realistically need in this scenario is to simplify the choice, boiling it down to a handful of alternatives: It's cognitively pretty much necessary to group candidates into parties.

But if we have like five parties, each treated as a single candidate, that's obviously insufficient for filling like 200 seats. After five seats you'd essentially run out. So how to fix that?

My proposal is just to rely on the clone independence of these methods: Pretend that there are, for each party, as many candidates as there are seats. So in our five party election with 200 seats to be distributed, we are looking at 200 * 5 = 1000 "candidates".

If you have the party choices

Cyan Magenta Orange Purple Yellow (or C M O P Y)

a ballot

O M Y P C

would mean that, to this voter, "Every Orange Party candidate beats every Magenta Party candidate beats every Yellow Party candidate beats every Purple Party candidate beats every Cyan Party candidate"

And then you just do the regular method of how Schulze STV works, simply pretending there is a massive amount of identical clones. - Because of the massive amount of symmetry, this is computationally likely much more feasible than trying to filter 1000 individual candidates down to 200 seats. And for a voter, it's much more reasonable to rank five parties than to rank a thousand people.

You lose nuance of course, but realistically, I don't see how that nuance wouldn't be lost anyways.

clearly this is fiber optic yarn based

I don't like renaming IRV to RCV as if that was the only RC method.

maybe make the bisons themselves roughly log scale (relative to where on the bar they land) too, making it look a little bit like herds in perspective?

Is this supposed to be Tideman Ranked Pairs? If so, the video actually isn't entirely clear to me as given. What are the wiggling ballots supposed to show? And perhaps it should be compared to how the actual result came about?

FWIW the Greens do have electoral reform on the agenda and they are rising fast atm

all Pearls come that way I'm pretty sure. The implication is, that Pink Pearl (Volleyball) had a certain unique personality and so when the whole tantrum happened and Pink Pearl was taken away, Pink wanted Pearl to be whoever she is, not giving any special instructions, so Pearl went default mode. But the other Pearls would also have been like that before the initial boot sequence

this feels lke an early styling thing. It's gonna improve. As of right now though, classic hands down. The animated versions lose way too much character and just feel like weird blobs

Robinson is better (but imperfect) about preserving area but it can't preserve relative distances either. I don't think any projection short of a literal globe can do that.

Regardless, it's a cool projection

Do your food islands drift? If so, what's going on is that those islands bounce around like that DVD player screen safer in a rectangular region, eventually leading to this.

Another way you can try: select the bottom vertex and snap the 3D cursor to it,

select the edge to be extruded (do not shift the 3D cursor) and set transform to 3D cursor,

extrude (E) and scale (S) - it should now scale from the bottom vertex at the center which happens to be extruding in the exact direction you indicated with those arrows

Yeah, sure.

In case of color vision you'd more or less have a "light sensing" module that can specify wavelength sensitivities, which could be copied and modified with different sensitivities, giving you the equivalent of the LMS receptors, and then you'd have some modules that do some low level preprocessing or whatever.

You just have a parametrized building block (a light sensitive cell) that can be copied many many times with various parametrizations (to build up a retina capable of detecting many different kinds of light)

Not saying that's precisely what Bibite Vision ought to be like, but that sort of idea (simple building blocks that can come in many forms) is what I have in mind.

What I don't want to see is like

- Level 1: Binary Vision (light/dark)

- Level 2: Monochrome vision (grayscale)

- Level 3: Color vision (three specific fixed color channels)

It makes sense that this level 3 is locked behind investing in Level 1 and Level 2, but it's far too inflexible a system. There shouldn't be fixed prefab levels (i.e. tiers) like that. Instead, there should be simple building blocks where tiers like this may be *emergent,* if that makes sense.