Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[WanderingWinder]

Argh. This is actually quite hard.

There is some real "power" in being allowed to create the poll, and the some more in "interpreting the results". I guess that's why politics is such a mess.
I really want to select the structure that is preferred by the most people.

I like your attempt to break it down Kirian, but I also agree with yed that it leads to false conclusions.

Yes, you definitely had some power when creating the poll, and also some when interpreting it. I think Kirian's interpretation is solid though, because the differences between medium (36) vs small (21) and no relegation (32) vs relegation (23) are so big.

If you like to know more about voting systems, check out this video for a better way to vote (the only disadvantage is that every voter has to rank the options from most preferred to least preferred, which is not really possible on this forum).

That's not the only disadvantage.... that system (like every other, if we want to be fair) has some serious problems.

[GeoLib]

Personally, I'm a fan of range voting (give each option a score from 0-9, highest average wins), which would be relatively easy to implement with the forum polling system.

Given that we're pretty sure that large is out of the picture, this could be done as four separate polls (small w/relegation, small w/out, medium with, medium w/out), each with 10 options (numbers 0-9). It would be trivial to calculate the average score for each option from these polls.

Obviously this can also be done over a range other than 0-9. Can you do separate polls on the same topic?

[soulnet]

Personally, I'm a fan of range voting (give each option a score from 0-9, highest average wins), which would be relatively easy to implement with the forum polling system.

You need to at least normalize the scores (divide the scores given by X by the sum of all scores given by X). And even then, I am not sure that represents anything. Theory of voting is hard.

[GeoLib]

Personally, I'm a fan of range voting (give each option a score from 0-9, highest average wins), which would be relatively easy to implement with the forum polling system.

You need to at least normalize the scores (divide the scores given by X by the sum of all scores given by X). And even then, I am not sure that represents anything. Theory of voting is hard.

No you don't (If I'm understanding what you're saying correctly). Your proposal of normalization means that each person has some amount of "vote" which they get to distribute amongst the various options. The point of range voting is that everyone ranks each option by how much they like it and we go with the one that has the highest average approval. It's like the approval voting as implemented in this poll except that there are more than just the binary options of approve and disapprove.

[soulnet]

Personally, I'm a fan of range voting (give each option a score from 0-9, highest average wins), which would be relatively easy to implement with the forum polling system.

You need to at least normalize the scores (divide the scores given by X by the sum of all scores given by X). And even then, I am not sure that represents anything. Theory of voting is hard.

No you don't (If I'm understanding what you're saying correctly). Your proposal of normalization means that each person has some amount of "vote" which they get to distribute amongst the various options. The point of range voting is that everyone ranks each option by how much they like it and we go with the one that has the highest average approval. It's like the approval voting as implemented in this poll except that there are more than just the binary options of approve and disapprove.

Suppose just 2 options. If you assign scores of 1 and 2 and I assign scores of 8 and 4, it basically means that we have opposite views. However, my vote influences the final result much more than yours. Even if you disagree with the model I propose, there are deeper problems: the incentive is not to give your honest opinion, for instance. E.g., if I prefer only slightly to have small groups, I might as well give scores far away from each other, such that my preference, thought not really strong, has strong influence on the outcome.

[GeoLib]

I agree that there's the potential for strategic voting, but your example doesn't really make any sense. Why would we ever use this system if there were only two options? Why would I give scores of 1 and 2? Why would you give scores of 8 and 4? None of these actions are logical.

Basically the idea is that everyone would give there most preferred option a 9, there least preferred a 0, and rank the other two somewhere in between based on how much they like them relative to the others.

[soulnet]

Basically the idea is that everyone would give there most preferred option a 9, there least preferred a 0, and rank the other two somewhere in between based on how much they like them relative to the others.

I just find no theory to support such system, and too complex not to offer theoretical guarantees. I prefer the simple voting or a complex system that guarantees no strategic voting is incentivized. I will stop arguing now, this is almost RSP material at this point.

[Voltaire]

At this point, I think an executive decision of something that's at least sorta-close to something that got a reasonable amount of votes is the best way to go. We could spend forever trying to find the exact consensus setup.

[-Stef-]

At this point, I think an executive decision of something that's at least sorta-close to something that got a reasonable amount of votes is the best way to go. We could spend forever trying to find the exact consensus setup.

Yes, I'll do that. Sorry I let it go for a few days but the last couple of days have been busy (Yesterday was my birthday )

My favorite option won the vote with 25-24-23, yet I'm going to disregard it. I think Kirian's point about the majority being against relegation matches does make sense, and it does indeed have some organisational problems we don't need. I guess I fail miserably as a dictator-pretending-to-be-a-democrat here.

I will set up a new vote between option 1 & 2, that will run for a week, and then I'll actually stick with the results, even when it's just a 1-vote-difference.
Signups will start at the moment Gokodom 3 moves into the knock-out phase.

...  If I were running this, I'd at least consider eight-player league where 1 and 2 promote, and 5-8 demote, to significantly increase churn.

No I'm sorry I really dislike this one, enough to not even want to vote on it. 4 people with unconditional demotion implies someone will know for sure he will demote 2-3 matches before the end. The remaining matches will still have a significant impact on others though, probably even on who ends up #1. Also because we don't count win-tie-loss, but the actual number of games won. Even when this player acts in good faith and plays on as good as he can, it's just a non-fun position to be in.

[Kirian]

...  If I were running this, I'd at least consider eight-player league where 1 and 2 promote, and 5-8 demote, to significantly increase churn.

No I'm sorry I really dislike this one, enough to not even want to vote on it. 4 people with unconditional demotion implies someone will know for sure he will demote 2-3 matches before the end. The remaining matches will still have a significant impact on others though, probably even on who ends up #1. Also because we don't count win-tie-loss, but the actual number of games won. Even when this player acts in good faith and plays on as good as he can, it's just a non-fun position to be in.

I hadn't thought of that, and you're totally right.

[soulnet]

...  If I were running this, I'd at least consider eight-player league where 1 and 2 promote, and 5-8 demote, to significantly increase churn.

No I'm sorry I really dislike this one, enough to not even want to vote on it. 4 people with unconditional demotion implies someone will know for sure he will demote 2-3 matches before the end. The remaining matches will still have a significant impact on others though, probably even on who ends up #1. Also because we don't count win-tie-loss, but the actual number of games won. Even when this player acts in good faith and plays on as good as he can, it's just a non-fun position to be in.

I hadn't thought of that, and you're totally right.

A good middle-ground could be deciding conditional demotion by some relative notion of the number of points obtained, instead of a direct match, which has additional complications. Something like, each #2 promotes, demoting a #5 or #6, if it obtains 2x or 3x or whatever proportion seems reasonable the number of points in the last season. Pair all #2s of the same level in decreasing order of points with #5s and $6s in the level above in increasing order of points.

If we end up with the medium size, this can even be introduced after some seasons.