AU mod: The Statue of Zeus

[alexman]

That would work too.

With a 8-turn interval, even if you build the 300-shield Wonder by 2150 BC (40 turns into the game), you wouldn't launch your attack with 10 units until 150 BC (80 turns later).

300 shields is the price of 10 Horsemen, so you might still be better off investing in 10 Horsemen instead of Ancient Cavalry in some rare cases when the tech rate is fast or when you don't get around building Wonders until late.

So how about it? 300 shields and 8-turn interval under consideration?

[nbarclay]

I still like my idea of having each panelist suggest his preference for weakening the wonder, having a vote among those, and having a run-off if none of those gets a majority. That would avoid creating a bias based on which of the ideas is chosen to put up for a vote first.

[ducki]

I see your point on that, Nathan, but also the fact that any given proposal can be voted down and returned to discussion is, IMO, a balancing factor in the case that any bias in the process might be introduced.

Creation by committee is always challenging, but I think someone has to step in and sift through the discussion to come up with a proposal to be voted on. With the panel, there's far less ... unilateral decision-making ... based on seeming concensus than previously and I'd rather someone(currently alexman has assumed/retained the mantle) regularly do that so we don't devolve into eternal discussion leading to too little action.

I don't necessarily agree with all of the ... filtered proposals ... placed under consideration, but I have to just trust in the panel to take into account everything discussed and vote for or against items keeping in mind the goal of the mod.

I think the open discussion period is where everyone suggests their preferences. We've gotta have one person pick items or everyone on the panel will vote for their favorite(their own proposal) and nothing passing.


Good lord! I've got to work on shortening these posts. I'm a madman when I don't want to work.

[nbarclay]

As long as a consensus seems to be emerging around one particular way of addressing an issue, I agree that having a single vote is a lot more streamlined. But when a lot of people still have a lot of different ideas about what approach is best, and the panel itself is divided, I think a multi-layer vote is better.

If we put one particular proposal up for a vote now, how should a panelist vote if he views the idea under consideration as the second best option available? Should he vote for it for fear that if it is rejected, what is ultimately accepted will be something he likes less? Or should he vote against it in the hope that the idea he likes better will be adopted?

An election/runoff approach isn't all that much more cumbersome, and it significantly increases the odds of ending up with the approach that the panel as a whole considers best. With how widely opinions are still divided on this issue, I think that's worthwhile. I'd rather not risk compromising the quality of the final decision just because we're in a hurry to settle the matter.

Nathan

[alexman]

This is how it currently works:

We vote for a proposal when we think it will improve the game from what is already in the mod.

An approved change doesn't end the process. If there is another proposal for consideration, the panel will vote for it if they consider that it's better than what's in place already.

[ducki]

Also, most changes are reviewed after the next game. Version 1.0 is a bit tougher, but once we get the first rev and play a game, we can see the changes in action and go back to changes that are less than optimal.

I'm all for democratization, but eventually some single person has to come out and say "let's vote on X". Otherwise, Firaxis will release the Official Patch and we'll still be working on getting an initial test version of the mod ready to play.

I'd like to play an AU game now with a ported AU:PtW judiciously combed over by the panel or even just alex with no discussion. Once we play, it'll be easier to see things that need change. If we never play, we'll never change anything.

[nbarclay]

After listening to all sides of the debate, the AU mod panel then formulates a specific proposal, and the the issue is marked "under consideration" for a period of one week. During that period, the community is given the opportunity to present arguments for and against the proposed change.

To me, this implies that deciding what specific proposal should put up for a vote is supposed to be a collective process by the panel as a whole, not a process in which one particular individual unilaterally says, "Let's vote on such-and-such." As long as there is a reasonably clear consensus by the panel regarding what particular change ought to be put up for a vote, there is no problem with one person saying, "Okay, this seems to be what we want, let's vote on it." But in the absence of such a consensus, and especially when it looks like there might be almost as many preferred solutions as there are panelists, I don't view having one person unilaterally decide what will be voted on first as following the process of "the panel" deciding what specific proposal will be voted on.

Also note that my proposal might well be faster than having a series of votes for everyone's preferred ideas individually, especially if panelists don't let a successful vote for one idea deter them from calling for a vote on something else they like better. If there are four or five different ideas preferred by one or more panelists (and my impression is that that is very likely the case), having a vote among them with run-offs as appropriate would require fewer votes than if each idea is voted on individually.

Nathan

[vulture]

How about the good old single transferable vote system, just to make life confusing; although it is actually less confusing than it sounds.

Say there are 4 options: A, B, C and D. Each voter lists the options in order of preference. You then go through a loop of discarding the option(s) with the lowest number of votes, and reallocating those votes according to their next preference.

e.g. 5 voters, 4 options. Each lists their order of preference as
1) ABCD
2) ACBD
3) BACD
4) BCAD
5) CBAD

So after the initial round, A has 2 votes, B has 2 votes, C has 1 vote and D has no votes.

So we delete option D, and re-allocate it's votes. There are none, so this doesn't take long. Then we delete option C, since it is the next least popular. That vote (number (5)) is reallocated. The second preference there is B, so 5's vote is added to B's total, giving A 2 votes and B 3. B wins.

The reason for doing this procedure rather than going for a simple majority is basically to compensate for a split of the protest vote. For example, suppose at the next US election, the candidates are Bush, Dean and Hillary Clinton standing as an independent Bush gets 40%, Dean and Clinton both get 30%. But all the Dean and Clinton supporters want Bush out. First past the post leaves him in. STV would mean that Hillary got eliminated, with her supporters votes going to their second choice candidate (which one assumes would be Dean - or there might be a 'even Bush is better than Dean' thing going on).

Back to the AU. We could either adopt the most popular choice after transferable voting (or poll between the two most popular choices), or use this as a method to select which option gets put to a yes/no vote first.

[alexman]

I think that system similar to that would work well for us, vulture. Thanks!

The biggest problem I have with such a system that doesn't involve comparing each proposal head-to-head against every other proposal, is that votes between similar proposals can get split, while a third (more unique) proposal might win even though it would not win head-to-head against either of the other two proposals.

But a single vote asking panel members to order the proposals will be sufficient to determine any head-to-head winner, although not exactly by the single transferable vote system described above.

[alexman]

Example of a difficult vote:
Four proposals, A, B, C, D
Five voters 1, 2, 3, 4, 5
A and B are similar and can be considered almost the same proposal.
C, D are unique.

Voter 1: ABCD
Voter 2: BACD
Voter 3: CBAD
Voter 4: DABC
Voter 5: CBAD

Note that A and B are always next to each other, as they are almost the same idea.

Under nbarclays system, C would win, even though more panelists prefer the A/B proposal over C.

Under the transferable vote system, A would win, even though most panelists prefer B over A.

But if we move down the list, comparing A to B, then the winner against C, and then the winner against D, (essentially what we have been doing) B would win, as I think it should.

[vulture]

It is worth bearing in mind "arrow's theorem" (IIRC), which amounts to there being no voting system which doesn't throw up pathological results once in a while, although some do better than others. Actually there is one which perfectly represents the voting population: one man, one vote (as done by Terry Pratchett: "The patrician was the Man. He had the Vote.")

The flaw in a simple vote is fairly obvious. The flaw in single transferable vote system is pretty much what Alexman says - there are combinations of cotes that leave the 'wrong' answer in place,

The flaw (or one of the flaws - most systems can fail in a variety of amusing ways) in Alexman's proposal is this:

3 voters, 4 options, voting in order of preference
1) ABCD
2) CDAB
3) BCAD

compare A to B: A comes ahead of B twice. Compare A to C: C beats A twice, so become the preferred choice. compare C to D: C always beats D. So we have a clear winner: C. Or do we? Compare C to B: B beats C twice.
So B beats C, C beats A, A beats B. Rock-paper-scissors scenario, where it is impossible to chose a winner.

Incidentally, using tiebreakers under STV, A, B and C all get one vote, so there is no lowest candiate to throw away (well, apart from D, but that doesn't change anything at this stage). Looking at the second preferences as a tiebreaker, B, C and D all get one vote. D is gone, so 1)'s vote goes to his third choice - C. C wins the tiebreak 2-1 (if there is a tie at this stage, you can repeat the process of course).

[alexman]

I knew I had to be missing something!

[sabrewolf]

or, you could give everyone a vote for each x from {+2, +1, 0, -1}. where 0 can be used more than once.
the higher average wins. at a tie, the one with the smaller standard devation wins (so +1+1 is better than +2+0).
next vote decides, how many things get included.

a different idea:
if you can reduce the choices to 3 from 4, eg.
A: SoZ up to 300
B: frequency down to 7
C: AC stat 2,2,2
D: remove +1 HP
then you could vote for combos, which would exclude the possibility that a favourable combo idea get's killed because each single ideas are too weak. and it also makes sure that not too much is done (imho having more than 2 of the points mentionned above would kill the wonder)

[alexman]

OK, please continue all discussion about the voting system in the main AU mod thread (follow my signature). Thanks!

[Harovan]

Another way to nerf it would be to let the cost and all stats as they are and to require the resource to be inside the city radius. So the humans reaction to switch all to a SoZ prebuild in a good core city as soon as he spots ivory wouldn't work. Sure, if he's extremely lucky, he could just have ivory in his inner core. But that's an exception. It will more likely be farther away, somewhere in the higher corruption area, and the human has to send a settler, to waste precious worker capacity to improve some crappy peripheral city and to build it the hard way, shield by shield. Helps AI concurrents, because as we all know that the AI will build it only after discovering Mathematics and not necessarily in a good core city, but in the first crappy city that screams for work and has ivory.