@Voss
I was wondering - could you publish the calculations so we all can see them raw? :)
Printable View
@Voss
I was wondering - could you publish the calculations so we all can see them raw? :)
Judging from the paper he sent, the motivation behind inventing the voting system was to try and completely focus on people's raw preferences. If you could "meta-abuse" the system with strategies like "oh I know this person is gonna get voted unless I put him super low so I'll try and put him at the bottom" that would ruin the whole point =P.
@OzyWho if you see the vote counting as a long algorithm it won't be helpful. If you want a good guess of who's going to be the chosen lynch, here is a decent guide to guess:
1) Pick out the 2 or 3 people that are high up on everyone's list. (Suppose these 3 people are players A, B and C)
2) Count how many people placed A higher than B or vice versa, how many people placed B higher than C, exc.
3a) Case 1:
If A was higher than B on more lists than B in relation to A. And A has the same relation with C:
Then A is lynched.
3b) Case 2:
Suppose we have an A>B and B>C and C>A situation.
RNG time.
Voss basically does exactly the above, except rather than with the top 3 he goes through every person and computes all the possible preferences. It's a huge, painstaking computation.
You can make the following inferences/deductions:
-In practice your list is often irrelevant after the first few entries.
-If 10/14 people put A>B that's no different from 9/14 people putting A>B. All that matters is more people put A higher than B.
-How "hard" those people score A higher than B (i.e. putting A first and B last) is irrelevant. All that matters is more people score A higher than B.
In practice it's quite difficult for Case 2 to occur, because people naturally tend to have a strong consensus of who should be lynched anyway. (mob mentality n shit). But here's an example for 5 players:
Suppose there are 5 players. A, B and C are suspect while X and Y are not. Suppose noone is bothering to list X and Y because they're so towny. Suppose they have the following votes:
A: B, C, A
B: A, C, B
C: B, A, C
Among the 3 suspects A, B and C there is a consensus to lynch B.
X: C, B, A
Y: A, C, B
However, Y is staunchly opposed to lynching B and puts B last. X is down for lynching B but would rather see C get lynched. We end up with the following scenario:
B>A 3/5
C>B 3/5
C>A 3/5
Now, this situation is rather unstable. If anyone were to flip their first two options around, the person who would get freshly placed at first would be immediately lynched. So we've ended up with a rather implausible situation. In practice one of A, B and C would certainly flip their first two picks around ASAP or risk being lynched.
Case 2 also allows for minor exploitation. For example, Y may townread C more than B but it's in Y's interests to "lie" and put B last anyway to force a Case 2. However, if I'm visualizing this correctly, you can only exploit the vote in the case of more than 2 suspects. If only two people are the main suspects shuffling your list around will not flip the vote. (and generally people only have the attention span to juggle 2 people at once LOL)
It'd be interesting to see if a stable albeit unlikely example of case 2 could be presented.
Oops that was long o.o
My IRL situation is over. Addressing this now.
Thoughts on game.
Thanks for everyone that participated. It game me such an interesting perspective and I also had fun talking to all of you contestants. If I were to play in FM, I'd have such better reads on all of you now. Watch out.
There were a couple of issues about OGC that were raised during the course of the game, but specifics excluded, what do people think about tagging people outside of thread to get them to post? Technically, as our site definition, this would be OGC, but reminding people to post in a game ups the quality of the game, so that's the plus. There's no question that this is OGC and I asked people not to do it in this game, mainly because I don't like changing rules midway through a game.
But the question I'm posing is should that specific instance continue to be banned?
@Distorted can you be more specific in the way that I implemented cvote that it wasn't easy to monitor live, and how it wasn't easy to keep track of the votes? I think, given the tools I had, that I accomplished this. It just sounds like you're trying to critique condorcet, while the implementation is the problem.
I do agree with yzb though, and whoever else mentioned it in thread. Using the red bolded text to indicate someone being scummy is a good visual way to pressure someone, and the implementation doesn't do this.
Just a last reiteration that I needed a creative way to parse post texts, because I didn't want to manually and error prone-ly calculate the posts, and I thought that posting every 6-8 hours what the counts were is not real time, and a bit taxing.
But.
If I could edit the site code to allow a for cvote to be used, I'd not have it wrapped in a spoiler, and I'd color code it similar to how it is on narrator.systeminplace.net/condorcet.html. It could look something like this:
Vote Standing
Distorted, Ratatoskr, Dutes, Rumox, Blinkskater, Efekann02, naz, Lenneth, Marshmallow Marsh
and you'd do something similar to what Ozy said
[cvote]Distorted, Ratatoskr, Dutes, Rumox, Blinkskater, Efekannn02, naz, Lenneth, Marshmallow Marsh[/cvote]
The above brings back that psychological 'scare' that normal vote brings.
Unless there's something else that I'm missing that vote does.
Thoughts on use cases
Here's the way I see big game pressuring go. Vote trains are built up. If you agree with one of the candidates, you help build the train. If you don't, you start your own. If you're like me, and like multiple trains because you want a lynch on someone scummy, (even if they're not your #1 choice), you have to keep swapping your votes on the train that has the least amount of pressure. Then you get called out for not supporting a lynch on a scummy person and get called a scum partner, devolving the chat.
Condorcet allows you to pressure multiple people at once, with less individual work. It also eliminates the need to be around for end of day, because even if things shift around, you'd still probably be supporting the next most scummy vote.
I'm not sure why people keep saying that this would be useful in bigger games. I think the reason why the benefit wasn't as apparent in this game was because after day 1, people were only trying to find 1 scum. Rat was outed Day 2, and everyone knew what was going to happen after the out. The correlation is there, but I'm not quite sure how to word it.
The cool thing about condorcet voting is even if the host doesn't want to use it, you can still collaborate with those that want to follow it by calculating the most scummy person using condorcet, and take into the account the vote preferences of those that want to use it, and just input the single vote preference of those that don't want to use it. Not commenting on how practical this would be, I'm just saying it's possible.
Rant on current vote system used on sc2mafia
Single majority vote has been a long time use on the forum. However, these are the critiques that I have on it:
Being around at EoD
It often requires people to be around at End of Day. This is a bit silly, because the whole point of forum mafia, is that you don't have to participate 100% of the time. You do it when you can. But I guess when you play, you have to make an effort to be around at the end.
Hammer vote stops the day.
There's actually a number of subrants with this. There's a meta to 'hammering' the last vote in majority that makes you scummy, which I think is silly. If the eliminated person is bad and you didn't have a chance to vote the person you're automatically slightly associated with the person (not the biggest complaint). However, it'd be more accurate if there was more than 51% that wanted the eliminated person dead to be able to express this. I don't see why this has to do with mafia, and I see this more as an 'easy way' to automatically end the day without host intervention.
But lastly, when games get down to unique 'final' scenarios, (like 2v1 no clear scenarios), you get (in my opinion) the unintended consequence of 'clears'. If I vote someone, and the other person hasn't instantly hammered, I know that either I've already lost the game, or I've essentially cleared a person from my perspective. I don't like this scenario because I don't really express who I think is mafia, I play a risk on me ending the game. The way these situations should end should be a clear consensus that 2 people think one person should be eliminated. Exactly how Efe, Dutes, and Distorted ended the game. Not a gimmicky trick, just straight convincing people.
Majority vs Plurality
Not really an issue this game because I made lynches mandatory, but I also dislike the fact that most people use 51% as the minimum determination of who gets eliminated. Going back to the being around for End of Day, if someone makes a good defense, and people need to try to switch to someone else at the end of day, but some people don't come back in time, then a lynch doesn't go through. Which sucks. Sure I can see arguments in Real Time mafia requiring majority over plurality, but in a game that doesn't require people to be on the site 100% of the time, this should be the standard. It punishes the players for things outside of their hands, gives evils an extra round of killing because town wasn't able to muster the votes on someone. It also kills the game because of the hopeless feeling town gets when a day ends because the previous day was wasted. I honestly don't see why people use majority over plurality.
Condorcet voting addresses pretty much all of these concerns. I also don't see how the system is actually town/mafia sided. What's mafia sided is the implementation of the thread lock hammer that's used because it's always been used.
I didn't host the game to bash on the way games are run. This was just a cool thought experiment. Not sure where I'm going to take this except to probably finish implementing it on the narrator *Shrug*
If people want to know how this is generally implemented, they're going to have to actually look it up. It's been explained by yzb here, and by me in the setup thread for this. It's not that difficult to understand, especially for people who play mafia for fun =P
The short answer that I've been giving is that it takes everyone's vote preferences, mashes them together and gives a list of who's most suspicious to the least. Like an average. If most people have person A in the tops of their lists, they'll generally be ranked higher.
I'm not sure how much of my words are making sense, but I further hope that no one reads too much negativity in what I'm saying. It's just a thought experiment/exploration, and I'm thankful for all the interest/comments people have made, and specially thankful for the players that experimented with it. <3
@Voss
Let's say this voting system would be implemented in the website as an option. Would the games that are choosing cvote, would they require to always have odd number of players alive, like you had in your game? (It was "long" ago when you made the decision, so I forgot everything about it, so I guess others did too and I figure I can just ask instead of searching back :p)
I will personally be using this voting system as a host, and I think I'm going to be using something similar as a player - as I said, I'm very indecisive, and having a hierarchical list of 'scummy' people is very useful to me. At the very least, it eliminates part of the confusion XD
Hmm... how would you go about to do it if it doesn't get implemented in the website?
The only way I can think of how to at least semi-automate the process is to save everyone's cvotes as CSV files, import the CSV files into excel or google sheets, and then do the calculations with formulas. I have never tried to do anything with CSV files, so not sure if it could work. But if it does then everyone could at least see current vote standings in the google sheet.
The only downsides (if it works) is that you would still have to manually input everyones cvotes AND retype names so every name is always written the same exact way (so the formulas could work) because people would write names differently.
I could see it as something doable, but waaaay too much work.
So my 2nd question to you is:
would you even try the voting system as a host if it doesn't get implemented in the website?
Me coding it so it's automated means the requirement of having odd players is relaxed. So, no it doesn't need odd players anymore.
As a player you can always use this vote system 'manually' as well! State your suspicious list and then go with the vote that's highest. Or something like that...
People could just use narrator.systeminplace.net/condorcet.html. I had it hard coded to point at a single thread, but it can easily be told to look at a particular thread.
I don't see such option there :(
I hope you don't mean to copy your code to ones own website or something like that. :P
Anyway. @Voss
Could you take a look at this and say if this is how the calculations were made? I made a Day1 example:
https://docs.google.com/spreadsheets...it?usp=sharing
It's chaotic and without comments so I think most wouldn't understand what's what there. But I think you can easily. My tip, if want to look into the formulas then it's easier to download as excel file and then "Evaluate Formula".
technically, but it doesn't handle ties, even numbered people, nor people not submitting a full vote list.