Provable randomness

[PokerFace]

1-(6/7 * 5/6 * 4/5 * 3/4) = 58%?

are you sure the 1- is necessary? shouldn't it be just 42%?

[Magua]

42% would be the chance that the assassin isn't lynched (6/7 chance they're not lynched D1, 5/6 chance they're not lynched D2, etc). 1 - 42% => 58% would be the chance the assassin is lynched.

[GreyICE]

Town on average in the early portions of the game lynches worse then randomly, in fact setups are generally created to be slightly town favouring which means dice tags are the optimal strategy if they're allowed.

Really?

A random lynch on a 10:3 should give you roughly a 23.1% chance of hitting scum. I'd wager that we're better than 1:4 on day 1 lynches.

[vettrock]

Town on average in the early portions of the game lynches worse then randomly, in fact setups are generally created to be slightly town favouring which means dice tags are the optimal strategy if they're allowed.

Really?

A random lynch on a 10:3 should give you roughly a 23.1% chance of hitting scum. I'd wager that we're better than 1:4 on day 1 lynches.

When I was looking at Newbie games, this definitely wasn't the case. While this is anecdotal, when I first started I looked at the last 10 newbie games, and there was zero D1 scum lynches. I think I had to go back to 15 before I found a D1 scum lynch.

[GreyICE]

From the fact it was 10:3, I'd like to point out I was discussing real games.

Newbie games are hives of scum and villainy, and while finding scum day 1 is easy, getting them lynched over the derpasaurus rex that is in the game (and typically in the SE slot) is hilariously difficult.

[GreyICE]

Okay, just checked the last 12 complete mini normals, and it was 7:5 town:scum

Not a large enough sample size obviously, but I'd be surprised if the town was below random (which would be about 9:3)

[Magua]

Any data set we use aside from newbies is going to be too small to be statistically significant, but there's the often-used statistic that of the 11 instances of 11:2 mountainous (40% EV) games run on mafiascum, town have only won one.

[GreyICE]

We've run 11:2s? I'm pretty sure the data set you're discussing is a 10:2, and 10:2s are brutally scumsided. 11:2s are theoretically balanced.

[Magua]

Mmmm. You're right that I may be thinking of the 10:2 era. But 10:2 still has an EV of 35%, which actual play was falling far, far, far short of.

[GreyICE]

Mmmm. You're right that I may be thinking of the 10:2 era. But 10:2 still has an EV of 35%, which actual play was falling far, far, far short of.

While I do think that 10:2 has a demotivating factor that makes it shit (it's a setup where no lynching day 1 without talking is correct for fucks sake) I'm not sure a record of 9-1 is so far outside that, statistically. Regardless, I'm pretty sure there's setup components that skew off strict statistics (otherwise this game would be pointless).

I do think that it's impossible to avoid a random lynch if people really want to. You can use pseudorandomness based on things like the timestamp of moderator votecounts, for instance.

[PokerFace]

The software I used to use to compile c# does not seem to want to work. So I did what I could math wise.

I figured if I was gonna find an error it would need to be a big one to send the 60% to below 50% and I did not find said big error. So your math is good

I did find it interesting that 6 player AITP was 50% ev with random lynching. I once believed 5 player was the most balanced version of AITP. Seems its actually 6

I designed something called WITP a while back. Can your code be applied to it to determine when random lynching gives each kingdom 33% odds and assassin 34%? There are 2 kingdoms in WITP so striving for 33%, 33%, 34% would be most balance I could get.

[vettrock]

The software I used to use to compile c# does not seem to want to work. So I did what I could math wise.

I figured if I was gonna find an error it would need to be a big one to send the 60% to below 50% and I did not find said big error. So your math is good

I did find it interesting that 6 player AITP was 50% ev with random lynching. I once believed 5 player was the most balanced version of AITP. Seems its actually 6

I designed something called WITP a while back. Can your code be applied to it to determine when random lynching gives each kingdom 33% odds and assassin 34%? There are 2 kingdoms in WITP so striving for 33%, 33%, 34% would be most balance I could get.

The wiki link is broken for WITP.

[PokerFace]

well fuck. try this

War in the Palace

[Magua]

Not sure that town in WitP benefits from a randomly determined lynch, because the odds of a random person being on your team are < 50%, unlike AitP where the odds of a random person being on your team are (x-1)/x.

Most strategies fall apart because town has an incentive to lie about their allegiance.

[PokerFace]

In playing the game it seems most balanced at either 7 or 9 total players. You can't have 8 players since you got 1 assassin, 2 kings, and 2-3 guards for each king (4-6 total). King A and King B are not on the same team and neither are their particular guards

Which do you think would be better balanced 7 or 9? I lean towards 7 as less players means less info for the assassin to find. In running the game I believe assassin has won 4 out of 6 times. One loss was with 7 and one was with 9. Three of the four assasin wins were with 9 and one of them was with 7

[MonkeyMan576]

I could see a situation where you are in lylo, 2 townie and 1 scum, player A(town) and B(scum) are veterans, but player C(town) is a newbie and would rather have the dice decide rather than be blamed for the loss. This would definately be against the spirit of the game.

[Soft-spoken]

here is an interesting example of why dice rolling is against your win-con

the overall odds of winning a 3-way with no clears as town is 1/3 assuming a random result. however, placing a random vote as town resolves this to a 1/4.

typically, people will withhold votes in a lylo until a lynch is decided on. in this case the 1/3 odds stay intact... but the second someone casts a vote without knowing for sure that it will be hammered, it decreases their odds.

they have a 50/50 chance of autoloss by voting town. if they miss this 50/50, they have the additional 50/50 of being crossed by scum and being hammered. the chances of casting a vote without knowing that it will be hammered are therefore 1/2*1/2 = 1/4

if everyone in town agrees to go along with a dice-roll, the odds are restored to 1/3, but this is unlikely, so the random vote decreases your odds of winning by a margin.

this also applies to scum, as a random vote turns a 2/3 into a 50/50

[PokerFace]

because of this thread, I have decided to run
Contagion in the Palace
during the next marathon day with 4 guards, 1 king, and 1 assassin. (6 players total)

Running it should provide some interesting observations. Feel free to join it during the july 2-5 marathon weekend http://forum.mafiascum.net/viewtopic.ph ... 6#p6956066

[Magua]

This should effectively limit discussion to zero outside day 1. And the assassin has even less means of protesting their lynch as it only take one player to eliminate the assassin with the disease. This version of AITP truly is a lottery.

What are you intending to get out of this?

[vettrock]

I think the rule, "No asking or prompting other players to commit to provable randomness" is suffient to prevent the "bad things" from happening.

[PokerFace]

This should effectively limit discussion to zero outside day 1. And the assassin has even less means of protesting their lynch as it only take one player to eliminate the assassin with the disease. This version of AITP truly is a lottery.

What are you intending to get out of this?

According to previous research on AITP with provable randomness, 6 players results in 50-50 odds for both sides

The first running of CITP was ruined by the mod revealing players' roles only

after

they had selected who they would infect next. The player's role should be revealed once the player is infected. This is

before

they select their infection target.

Yes, the game has a lottery element, but it may not be "directly" hindered or enforced. During the game the

thread will not be locked at anytime

. While a player is selecting their target, anyone is allowed to post. Will you help town guards, as they select who to infect? Will you try to confuse the Assassin, as he selects who to infect?

I'm basically running a game where provable randomness results in 50-50 odds. While allowing players to attempt wifom should they perceive it can better their odds. I plan to study if wifom is used in the heat of the moment. And how it is used in a heat that is generated faster than in normal AITP.

I am running this game during marathon weekend as its lottery element makes it unsuited for the main forums and I feel the heat/drama will increase in a marathon format without offending the mind sets of a player. Players get invested in month long games. Peoples feeling get invested in month long games. Peoples feeling get hurt in month long games. Something that is a lottery should not be an investment that could hurt feelings

I plan to update that wiki page shortly as its initial draft lacks some of the information given here

[Magua]

@vettrock:

That is bad for the same reason every single "Here's a role, you can't claim anything about your role or you'll be modkilled" role is terrible. It's very, very difficult to have a rule based on subjective interpretation.

@Pokerface:

Good luck.

[PokerFace]

I edited http://forum.mafiascum.net/viewtopic.ph ... 5#p6959915

do you mean that sarcastically or are you actually curious, Magua?

[Magua]

A little of column A, a little of column B?

I don't think AitP is a good setup or worthy of being played as-is (for the same reason I don't think 8 town:1 SK would be a good setup), but you're running an experiment and I can respect that and don't want to piss all over it.

[PokerFace]

Well put

I have updated the wiki to better represent the purpose of CITP without dismembering its integrity/balance as much as it had previously been dismembered