Ya I get it. But they might be wrong. And I'm not going to claim.
-
Welcome to Smogon! Take a moment to read the Introduction to Smogon for a run-down on everything Smogon, and make sure you take some time to read the global rules.
FEDS VS SMUGGLERS - GAME OVER, FED VICTORY
- Thread starter internet
- Start date
TL;DR: Game is broken, the only way smugglers lose is if we allow the feds to coordinate, and if everyone stops posting, the only way they can coordinate is by outing themselves (in which case I have a second plan as well)
To build some credibility, I'll walk you through the exact kind of logic I used but will leave the code out to make it harder for scum to replicate (although tbh this took wayyyyy less time than I thought it would)
-14 players in game, 10 smugglers, 4 fed. WLOG player 0 is the person who started with the goods, player 1 is the person who has it now.
-To simplify matters, ties at the end of cycle 10 end in a fed win (everything is super smuggler sided anyway)
-rules were pretty easy to implement.
I am writing this with the thought in mind that feds will read this as well.
We have three main questions:
Question 1: Are smugglers favored to win if everyone acts completely randomly and cease communication
Question 2: Is acting non-randomly and based on reads better than acting randomly
Question 3: How robust is a smuggler win to scum strategies
Question 1: Are smugglers favored to win if everyone acts completely randomly and cease communication
1: Yes --
If the feds cannot coordinate and pick at random, from this current position, smugglers win roughly 94% of the time
Smuggler wins: 9380
Fed wins: 620
Question 2: Is acting non-randomly and based on reads better than acting randomly
2. SURPRISINGLY, YES
Essentially I create a rank ordered list of the players in the game, with 1 being the most townread and 14 being the least. We can imagine that on this list, four positions are occupied by smugglers. If our reads are perfect, then this is positions 11-14.
We can assign everyone probabilities of being handed the goods. A conservative and smuggler-friendly way to approach this is to let the top person get 14/105 chance of being handed the goods on any given night, 13/105, and so on. This is actually much flatter than we expect the actual numbers to look -- in reality we'd be looking at a 20% of getting the goods when you're townread. The flatter this distribution is, the more "friendly" this simulation is to acting on reads.
If we have perfect reads, then this is actually slightly WORSE than everyone picking randomly. The reason why this is worse is because it is now more likely that scum actually get two people to inspect the same person. The accuracy exponentially deteriorates the worse our readslist gets.
Question 3: How robust is a smuggler win to scum strategies
This is where I won't be sharing all of my results -- not because there is even a single winning scum strategy, but because there are some that push the game closer to even.
To give an example of this, suppose scum instead of doing 4 random picks with >2 getting inspections, gets 1 inspect per round. In that case our best strategy is to STILL act randomly. This gives us a 80% chance of winning
Smuggler wins: 7957
Fed wins: 2043
Realistically this is probably where I expect our advantage to be, and for me I think that's good enough.
Why is it so important for us to stop posting?
Because the win equity and the essence of this game boils down to whether or not scum can get on the same page consistently. There's lots of ways they can do this by posting, since they know the previous inspects from each night. However, if NOBODY POSTS, AT ALL, then there is NO MEDIUM BY WHICH THEY CAN COMMUNICATE EVEN CODED MESSAGES.
The only people who CAN post are CLEARED TOWNIES because we know they are not sending coded messages!
Why is it so important that we act randomly -- can i please give fruit to my townreads?
No! If you are more likely to give fruit to townreads, or less likely to give fruit to cleared/revealed players, and scum know this, then that /hurts/ our winning chances because everyone has a decent idea of where townreads are.
The best thing to do is to act totally randomly so to dodge the scum inspect.
WHAT HAPPENS NOW?
I would like feedback on this plan of action today, I don't think allowing scum to coordinate one day is too hurtful.
Starting next day, no posts in thread by anyone other than cleared townies (myself, cel, + possibly whoever we vote today)
IF TOWNIES AGREE TO NOT POST, THEN SCUM CANNOT POST, THEY CANNOT COORDINATE, WE WILL WIN THE GAME at /LEAST/ 80% of the time.
I have some plans for when scum decide to out themselves and more simulations to back that up.
For today, my criteria for voting is someone who will not be able to lead and coordinate scum if they are outed and who might be a good person to bounce ideas off of if they are town: this user is VOTE CAMRYN
To build some credibility, I'll walk you through the exact kind of logic I used but will leave the code out to make it harder for scum to replicate (although tbh this took wayyyyy less time than I thought it would)
-14 players in game, 10 smugglers, 4 fed. WLOG player 0 is the person who started with the goods, player 1 is the person who has it now.
-To simplify matters, ties at the end of cycle 10 end in a fed win (everything is super smuggler sided anyway)
-rules were pretty easy to implement.
I am writing this with the thought in mind that feds will read this as well.
We have three main questions:
Question 1: Are smugglers favored to win if everyone acts completely randomly and cease communication
Question 2: Is acting non-randomly and based on reads better than acting randomly
Question 3: How robust is a smuggler win to scum strategies
Question 1: Are smugglers favored to win if everyone acts completely randomly and cease communication
1: Yes --
If the feds cannot coordinate and pick at random, from this current position, smugglers win roughly 94% of the time
Smuggler wins: 9380
Fed wins: 620
Question 2: Is acting non-randomly and based on reads better than acting randomly
2. SURPRISINGLY, YES
Essentially I create a rank ordered list of the players in the game, with 1 being the most townread and 14 being the least. We can imagine that on this list, four positions are occupied by smugglers. If our reads are perfect, then this is positions 11-14.
We can assign everyone probabilities of being handed the goods. A conservative and smuggler-friendly way to approach this is to let the top person get 14/105 chance of being handed the goods on any given night, 13/105, and so on. This is actually much flatter than we expect the actual numbers to look -- in reality we'd be looking at a 20% of getting the goods when you're townread. The flatter this distribution is, the more "friendly" this simulation is to acting on reads.
If we have perfect reads, then this is actually slightly WORSE than everyone picking randomly. The reason why this is worse is because it is now more likely that scum actually get two people to inspect the same person. The accuracy exponentially deteriorates the worse our readslist gets.
Question 3: How robust is a smuggler win to scum strategies
This is where I won't be sharing all of my results -- not because there is even a single winning scum strategy, but because there are some that push the game closer to even.
To give an example of this, suppose scum instead of doing 4 random picks with >2 getting inspections, gets 1 inspect per round. In that case our best strategy is to STILL act randomly. This gives us a 80% chance of winning
Smuggler wins: 7957
Fed wins: 2043
Realistically this is probably where I expect our advantage to be, and for me I think that's good enough.
Why is it so important for us to stop posting?
Because the win equity and the essence of this game boils down to whether or not scum can get on the same page consistently. There's lots of ways they can do this by posting, since they know the previous inspects from each night. However, if NOBODY POSTS, AT ALL, then there is NO MEDIUM BY WHICH THEY CAN COMMUNICATE EVEN CODED MESSAGES.
The only people who CAN post are CLEARED TOWNIES because we know they are not sending coded messages!
Why is it so important that we act randomly -- can i please give fruit to my townreads?
No! If you are more likely to give fruit to townreads, or less likely to give fruit to cleared/revealed players, and scum know this, then that /hurts/ our winning chances because everyone has a decent idea of where townreads are.
The best thing to do is to act totally randomly so to dodge the scum inspect.
WHAT HAPPENS NOW?
I would like feedback on this plan of action today, I don't think allowing scum to coordinate one day is too hurtful.
Starting next day, no posts in thread by anyone other than cleared townies (myself, cel, + possibly whoever we vote today)
IF TOWNIES AGREE TO NOT POST, THEN SCUM CANNOT POST, THEY CANNOT COORDINATE, WE WILL WIN THE GAME at /LEAST/ 80% of the time.
I have some plans for when scum decide to out themselves and more simulations to back that up.
For today, my criteria for voting is someone who will not be able to lead and coordinate scum if they are outed and who might be a good person to bounce ideas off of if they are town: this user is VOTE CAMRYN
thrilled that when someone goes "hey can i run this game" now i also have to consider getting a mathematics degree to make sure that the game can't be exploited by those luigi does nothing and wins video memes to avoid another problem in the circus maximus game schedule. im so thrilled that we have another week+ of hearing "the best winning move for this social game is to not play." because hal did the math on "what if we intentionally tried to break a fun game in order to not have to do social stuff in a social game" and then decided to unironically encourage people who signed up to play a game to not play
Gunboat exists and it is actually quite fun, we should host it sometime
this phase of the game only will last until scum realize they need someone to step up and claim and then it's up to them on what good idea they have
"forcing the optimal move as being anti-social seems bad for a social gaming community"
"no you don't get it, the optimal move is being anti-social"
buddy you and i aren't disagreeing, i am telling you that i am more concerned w/ the concept of anti-socialing a social game than i am getting the win
"no you don't get it, the optimal move is being anti-social"
buddy you and i aren't disagreeing, i am telling you that i am more concerned w/ the concept of anti-socialing a social game than i am getting the win
kind of crazy how quickly you've switched between this and "let's social game"
well that’s her opinion - she’s also not advocating for nobody to talk. She’s just saying that there isn’t too much of a marked difference
She switched opinions… So what??
also I’m gonna be honest here I would indeed like to win this game by following Hal’s plan, but I signed up for a social game and not a game of luck
camryn has to put it bluntly been a non-factor this game, we can find a better person to bounce ideas off of - from this PL, I think CK / bluedoom are at a level where they wouldn't be great at coordinating votes as well as being alright to pop in and bounce ideas off of as you mentioned