[EM] Updates on 3-2-1 and GOLD (minor rule adjustments)

[Jameson Quinn]

Since creating 3-2-1 voting <http://wiki.electorama.com/wiki/3-2-1_voting>
and GOLD voting
<http://wiki.electorama.com/wiki/Geographic_Open_List/Delegated_(GOLD)_voting>,
I've continued to work on refining the rules. I'm arbitrarily deciding that
it's time to send an update now, because they're getting pretty stable.
Future changes are still possible, but I think that on each system the
chances are better than even that the only changes from here on will be
simplifying the wording, not substantive differences.

On 3-2-1, here are the rules in addition to the basic "3 most good, 2 least
bad, 1 more preferred" dynamic:

- Optional delegation: if you vote just one candidate as "good", any
blank ratings you leave will be filled in by that candidate's predeclared
ratings. Predeclarations default to "bad".
- 3rd semifinalist: the third semifinalist must...
- ...not be of the same party as the first two semifinalists (if
there are party labels). If this would happen, just skip to the next
highest "good" as the third semifinalist.
- This means a party can't win by running 3 clones against a
divided field. There could still in theory be "clone
parties", but it seems
unlikely that over 1/3 of voters would go along with such naked
strategizing.
- ...have at least half as many "good" ratings as the top
semifinalist. If this would happen, just take the top two as finalists,
skipping the second "2 least bad" round.
- This ensures that you can't win just by being a "stealth
centrist" with little direct support.

On GOLD, here are the changes from the original version:

- Pre-eliminations
- The top candidate in each riding, based on local votes alone, is
never eliminated.
- The second candidate in each riding, counting local votes only, is
eliminated only if their local votes are fewer than half those of the top.
- Others are eliminated by default, surviving only if their local
votes are more than half those of the top AND their total direct votes
(including non-local write-ins) are more than those of the top local
candidate. (For this rule, "top" is counted by local votes only,
but "those
of" includes non-local votes.)
- STV elimination order: eliminate in order of who's furthest behind in
their riding
- If a candidate's current full tally is 1000 votes (including
everything: local votes, direct write-ins, and transferred
votes), and the
top full tally of any remaining candidate in their riding is 2000, then
they are 1000 behind in their riding.

These rule adjustments for GOLD work to ensure that the voters within a
riding have a significant say in who represents them, and out-of-riding
votes matter only at the margin.

I've also been thinking about rule adjustments for 3RD (the nonpartisan
proportional system that's a cross between 321 and GOLD). That was part of
why I started the other thread on proportional summability. But I'm not
really satisfied that that system is "done" at the level of 321 and GOLD,
so I'm not going to talk about the current version of those rules here.

[Jameson Quinn]

This raises a larger question: should the inventor of a voting method be
able to change the rules without changing the name? Once a proposal is
public, who owns it — the inventor or the public?

I don't think the answer is always clear-cut. If I wanted to redefine 3-2-1
to be something totally different, that wouldn't be OK. If somebody else
wanted to use the name 3-2-1 for something substantially the same, but with
minor rule changes, I might talk or argue with them about whether their
rule changes were a good idea, but I think I'd probably still let them use
the name.

But for smaller rule changes, I think it's OK for the inventor of a method
to make adjustments.

Post by Jameson Quinn
Since creating 3-2-1 voting <http://wiki.electorama.com/wiki/3-2-1_voting>
and GOLD voting
<http://wiki.electorama.com/wiki/Geographic_Open_List/Delegated_(GOLD)_voting>,
I've continued to work on refining the rules. I'm arbitrarily deciding that
it's time to send an update now, because they're getting pretty stable.
Future changes are still possible, but I think that on each system the
chances are better than even that the only changes from here on will be
simplifying the wording, not substantive differences.
On 3-2-1, here are the rules in addition to the basic "3 most good, 2
- Optional delegation: if you vote just one candidate as "good", any
blank ratings you leave will be filled in by that candidate's predeclared
ratings. Predeclarations default to "bad".
- 3rd semifinalist: the third semifinalist must...
- ...not be of the same party as the first two semifinalists (if
there are party labels). If this would happen, just skip to the next
highest "good" as the third semifinalist.
- This means a party can't win by running 3 clones against a
divided field. There could still in theory be "clone parties", but it seems
unlikely that over 1/3 of voters would go along with such naked
strategizing.
- ...have at least half as many "good" ratings as the top
semifinalist. If this would happen, just take the top two as finalists,
skipping the second "2 least bad" round.
- This ensures that you can't win just by being a "stealth
centrist" with little direct support.
- Pre-eliminations
- The top candidate in each riding, based on local votes alone, is
never eliminated.
- The second candidate in each riding, counting local votes only,
is eliminated only if their local votes are fewer than half those of the
top.
- Others are eliminated by default, surviving only if their local
votes are more than half those of the top AND their total direct votes
(including non-local write-ins) are more than those of the top local
candidate. (For this rule, "top" is counted by local votes only, but "those
of" includes non-local votes.)
- STV elimination order: eliminate in order of who's furthest behind
in their riding
- If a candidate's current full tally is 1000 votes (including
everything: local votes, direct write-ins, and transferred votes), and the
top full tally of any remaining candidate in their riding is 2000, then
they are 1000 behind in their riding.
These rule adjustments for GOLD work to ensure that the voters within a
riding have a significant say in who represents them, and out-of-riding
votes matter only at the margin.
I've also been thinking about rule adjustments for 3RD (the nonpartisan
proportional system that's a cross between 321 and GOLD). That was part of
why I started the other thread on proportional summability. But I'm not
really satisfied that that system is "done" at the level of 321 and GOLD,
so I'm not going to talk about the current version of those rules here.

[Toby Pereira]

Minor changes can be seen as variants of the same method. For example, someone might define a Condorcet method using margins, and someone else might prefer the winning votes version. But you wouldn't give it a different name. You'd call it x with winning votes. But then I suppose that doesn't answer the question of whether the inventor or anyone else can change the "canonical" version - the version where you don't have to specify anything else about it.


From: Jameson Quinn <***@gmail.com>
To: electionsciencefoundation <***@googlegroups.com>; EM <election-***@lists.electorama.com>
Sent: Saturday, 10 June 2017, 17:04
Subject: Re: [EM] Updates on 3-2-1 and GOLD (minor rule adjustments)

This raises a larger question: should the inventor of a voting method be able to change the rules without changing the name? Once a proposal is public, who owns it — the inventor or the public?
I don't think the answer is always clear-cut. If I wanted to redefine 3-2-1 to be something totally different, that wouldn't be OK. If somebody else wanted to use the name 3-2-1 for something substantially the same, but with minor rule changes, I might talk or argue with them about whether their rule changes were a good idea, but I think I'd probably still let them use the name.
But for smaller rule changes, I think it's OK for the inventor of a method to make adjustments. 
2017-06-10 11:14 GMT-04:00 Jameson Quinn <***@gmail.com>:

Since creating 3-2-1 voting and GOLD voting, I've continued to work on refining the rules. I'm arbitrarily deciding that it's time to send an update now, because they're getting pretty stable. Future changes are still possible, but I think that on each system the chances are better than even that the only changes from here on will be simplifying the wording, not substantive differences.
On 3-2-1, here are the rules in addition to the basic "3 most good, 2 least bad, 1 more preferred" dynamic:
- Optional delegation: if you vote just one candidate as "good", any blank ratings you leave will be filled in by that candidate's predeclared ratings. Predeclarations default to "bad".
- 3rd semifinalist: the third semifinalist must...

- ...not be of the same party as the first two semifinalists (if there are party labels). If this would happen, just skip to the next highest "good" as the third semifinalist.

- This means a party can't win by running 3 clones against a divided field. There could still in theory be "clone parties", but it seems unlikely that over 1/3 of voters would go along with such naked strategizing.

- ...have at least half as many "good" ratings as the top semifinalist. If this would happen, just take the top two as finalists, skipping the second "2 least bad" round.

- This ensures that you can't win just by being a "stealth centrist" with little direct support.
On GOLD, here are the changes from the original version:
- Pre-eliminations

- The top candidate in each riding, based on local votes alone, is never eliminated.
- The second candidate in each riding, counting local votes only, is eliminated only if their local votes are fewer than half those of the top.
- Others are eliminated by default, surviving only if their local votes are more than half those of the top AND their total direct votes (including non-local write-ins) are more than those of the top local candidate. (For this rule, "top" is counted by local votes only, but "those of" includes non-local votes.)

- STV elimination order: eliminate in order of who's furthest behind in their riding

- If a candidate's current full tally is 1000 votes (including everything: local votes, direct write-ins, and transferred votes), and the top full tally of any remaining candidate in their riding is 2000, then they are 1000 behind in their riding.
These rule adjustments for GOLD work to ensure that the voters within a riding have a significant say in who represents them, and out-of-riding votes matter only at the margin.
I've also been thinking about rule adjustments for 3RD (the nonpartisan proportional system that's a cross between 321 and GOLD). That was part of why I started the other thread on proportional summability. But I'm not really satisfied that that system is "done" at the level of 321 and GOLD, so I'm not going to talk about the current version of those rules here.

----
Election-Methods mailing list - see http://electorama.com/em for list info

[Ted Stern]

IMO the 3-2-1 methodology of 3-best, 2-least-worst, 1-most-preferred is not
just a specific method, but a filter that could be applied to other
methods. In the spirit of the original 3-2-1, the initial method should
satisfy the favorite betrayal criterion (FBC).

I currently like the idea of using the 3-2-1 filter with Majority Judgment
as the initial "3-best" method, since 3-2-1 eliminates (at least
practically) the participation criterion issue with MJ.

On Sun, Jun 11, 2017 at 6:27 AM, 'Steve Cobb' via The Center for Election

If the name encapsulated a description of what the voting method is, the
name would change naturally with the method. When the name is the author's
name, or something cute like 3-2-1 or IRV, you'd append a clarification for
minor changes. If Toby came up with a significantly different and clever
variation of 3-2-1, maybe you'd call it Quinn-Pereira Voting.
About ownership... does Warren own Score Voting? Maybe 3-2-1 should be
called Three-Value Score Voting with Three-Step Tally.

Post by Toby Pereira
Minor changes can be seen as variants of the same method. For example,
someone might define a Condorcet method using margins, and someone else
might prefer the winning votes version. But you wouldn't give it a
different name. You'd call it x with winning votes. But then I suppose that
doesn't answer the question of whether the inventor or anyone else can
change the "canonical" version - the version where you don't have to
specify anything else about it.
------------------------------
*From:* Jameson Quinn
*To:* electionsciencefoundation; EM
*Sent:* Saturday, 10 June 2017, 17:04
*Subject:* Re: [EM] Updates on 3-2-1 and GOLD (minor rule adjustments)
This raises a larger question: should the inventor of a voting method be
able to change the rules without changing the name? Once a proposal is
public, who owns it — the inventor or the public?
I don't think the answer is always clear-cut. If I wanted to redefine
3-2-1 to be something totally different, that wouldn't be OK. If somebody
else wanted to use the name 3-2-1 for something substantially the same, but
with minor rule changes, I might talk or argue with them about whether
their rule changes were a good idea, but I think I'd probably still let
them use the name.
But for smaller rule changes, I think it's OK for the inventor of a
method to make adjustments.
Since creating 3-2-1 voting
<http://wiki.electorama.com/wiki/3-2-1_voting> and GOLD voting
<http://wiki.electorama.com/wiki/Geographic_Open_List/Delegated_(GOLD)_voting>,
I've continued to work on refining the rules. I'm arbitrarily deciding that
it's time to send an update now, because they're getting pretty stable.
Future changes are still possible, but I think that on each system the
chances are better than even that the only changes from here on will be
simplifying the wording, not substantive differences.
On 3-2-1, here are the rules in addition to the basic "3 most good, 2
- Optional delegation: if you vote just one candidate as "good", any
blank ratings you leave will be filled in by that candidate's predeclared
ratings. Predeclarations default to "bad".
- 3rd semifinalist: the third semifinalist must...
- ...not be of the same party as the first two semifinalists (if
there are party labels). If this would happen, just skip to the next
highest "good" as the third semifinalist.
- This means a party can't win by running 3 clones against a
divided field. There could still in theory be "clone parties", but it seems
unlikely that over 1/3 of voters would go along with such naked
strategizing.
- ...have at least half as many "good" ratings as the top
semifinalist. If this would happen, just take the top two as finalists,
skipping the second "2 least bad" round.
- This ensures that you can't win just by being a "stealth
centrist" with little direct support.
- Pre-eliminations
- The top candidate in each riding, based on local votes alone, is
never eliminated.
- The second candidate in each riding, counting local votes only,
is eliminated only if their local votes are fewer than half those of the
top.
- Others are eliminated by default, surviving only if their local
votes are more than half those of the top AND their total direct votes
(including non-local write-ins) are more than those of the top local
candidate. (For this rule, "top" is counted by local votes only, but "those
of" includes non-local votes.)
- STV elimination order: eliminate in order of who's furthest behind
in their riding
- If a candidate's current full tally is 1000 votes (including
everything: local votes, direct write-ins, and transferred votes), and the
top full tally of any remaining candidate in their riding is 2000, then
they are 1000 behind in their riding.
These rule adjustments for GOLD work to ensure that the voters within a
riding have a significant say in who represents them, and out-of-riding
votes matter only at the margin.
I've also been thinking about rule adjustments for 3RD (the nonpartisan
proportional system that's a cross between 321 and GOLD). That was part of
why I started the other thread on proportional summability. But I'm not
really satisfied that that system is "done" at the level of 321 and GOLD,
so I'm not going to talk about the current version of those rules here.
----
Election-Methods mailing list - see http://electorama.com/em for list info
--

You received this message because you are subscribed to the Google Groups
"The Center for Election Science" group.
To unsubscribe from this group and stop receiving emails from it, send an
For more options, visit https://groups.google.com/d/optout.

[Jameson Quinn]

I agree that MJ3-2-1 would be a good method. (Minor point: I'd use
graduated majority judgment instead of MJ, because it gives a simple
real-number "graduated median" which can be used for tiebreakers, while the
MJ tiebreaker was optimized for compliance proofs).

Still, as a practical reform proposal, I don't want to pollute 3-2-1's
relative simplicity.

Post by Ted Stern
IMO the 3-2-1 methodology of 3-best, 2-least-worst, 1-most-preferred is
not just a specific method, but a filter that could be applied to other
methods. In the spirit of the original 3-2-1, the initial method should
satisfy the favorite betrayal criterion (FBC).
I currently like the idea of using the 3-2-1 filter with Majority Judgment
as the initial "3-best" method, since 3-2-1 eliminates (at least
practically) the participation criterion issue with MJ.
On Sun, Jun 11, 2017 at 6:27 AM, 'Steve Cobb' via The Center for Election

If the name encapsulated a description of what the voting method is, the
name would change naturally with the method. When the name is the author's
name, or something cute like 3-2-1 or IRV, you'd append a clarification for
minor changes. If Toby came up with a significantly different and clever
variation of 3-2-1, maybe you'd call it Quinn-Pereira Voting.
About ownership... does Warren own Score Voting? Maybe 3-2-1 should be
called Three-Value Score Voting with Three-Step Tally.

Post by Toby Pereira
Minor changes can be seen as variants of the same method. For example,
someone might define a Condorcet method using margins, and someone else
might prefer the winning votes version. But you wouldn't give it a
different name. You'd call it x with winning votes. But then I suppose that
doesn't answer the question of whether the inventor or anyone else can
change the "canonical" version - the version where you don't have to
specify anything else about it.
------------------------------
*From:* Jameson Quinn
*To:* electionsciencefoundation; EM
*Sent:* Saturday, 10 June 2017, 17:04
*Subject:* Re: [EM] Updates on 3-2-1 and GOLD (minor rule adjustments)
This raises a larger question: should the inventor of a voting method be
able to change the rules without changing the name? Once a proposal is
public, who owns it — the inventor or the public?
I don't think the answer is always clear-cut. If I wanted to redefine
3-2-1 to be something totally different, that wouldn't be OK. If somebody
else wanted to use the name 3-2-1 for something substantially the same, but
with minor rule changes, I might talk or argue with them about whether
their rule changes were a good idea, but I think I'd probably still let
them use the name.
But for smaller rule changes, I think it's OK for the inventor of a
method to make adjustments.
Since creating 3-2-1 voting
<http://wiki.electorama.com/wiki/3-2-1_voting> and GOLD voting
<http://wiki.electorama.com/wiki/Geographic_Open_List/Delegated_(GOLD)_voting>,
I've continued to work on refining the rules. I'm arbitrarily deciding that
it's time to send an update now, because they're getting pretty stable.
Future changes are still possible, but I think that on each system the
chances are better than even that the only changes from here on will be
simplifying the wording, not substantive differences.
On 3-2-1, here are the rules in addition to the basic "3 most good, 2
- Optional delegation: if you vote just one candidate as "good", any
blank ratings you leave will be filled in by that candidate's predeclared
ratings. Predeclarations default to "bad".
- 3rd semifinalist: the third semifinalist must...
- ...not be of the same party as the first two semifinalists (if
there are party labels). If this would happen, just skip to the next
highest "good" as the third semifinalist.
- This means a party can't win by running 3 clones against a
divided field. There could still in theory be "clone parties", but it seems
unlikely that over 1/3 of voters would go along with such naked
strategizing.
- ...have at least half as many "good" ratings as the top
semifinalist. If this would happen, just take the top two as finalists,
skipping the second "2 least bad" round.
- This ensures that you can't win just by being a "stealth
centrist" with little direct support.
- Pre-eliminations
- The top candidate in each riding, based on local votes alone,
is never eliminated.
- The second candidate in each riding, counting local votes only,
is eliminated only if their local votes are fewer than half those of the
top.
- Others are eliminated by default, surviving only if their local
votes are more than half those of the top AND their total direct votes
(including non-local write-ins) are more than those of the top local
candidate. (For this rule, "top" is counted by local votes only, but "those
of" includes non-local votes.)
- STV elimination order: eliminate in order of who's furthest behind
in their riding
- If a candidate's current full tally is 1000 votes (including
everything: local votes, direct write-ins, and transferred votes), and the
top full tally of any remaining candidate in their riding is 2000, then
they are 1000 behind in their riding.
These rule adjustments for GOLD work to ensure that the voters within a
riding have a significant say in who represents them, and out-of-riding
votes matter only at the margin.
I've also been thinking about rule adjustments for 3RD (the nonpartisan
proportional system that's a cross between 321 and GOLD). That was part of
why I started the other thread on proportional summability. But I'm not
really satisfied that that system is "done" at the level of 321 and GOLD,
so I'm not going to talk about the current version of those rules here.
----
Election-Methods mailing list - see http://electorama.com/em for list info
--

You received this message because you are subscribed to the Google Groups
"The Center for Election Science" group.
To unsubscribe from this group and stop receiving emails from it, send an
For more options, visit https://groups.google.com/d/optout.