Civilization Mafia - Town Wins!

[EmpTyger]

Thok:

<snip>EmpTyger, we do learn something from random votes because we see who tries to follow the bandwagons, and what bandwagons don't take off. Go read Hospital mafia in normal. Because nearly everybody in that game is a doctor, a no-lynch strategy is useful. However, most of the scum in that game were caught because of discussion arising from random voting.

Um... what??? Firstly, so far every day in Hospital mafia has had a lynch, so the example is moot. Secondly, on whatever day we decide to stop no-lynching will we will have the same opportunity to catch scum. Thirdly, if the optimal play is a no-lynch, there shouldn’t be any bandwagons. [And finally, I’d rather not comment on what may or may not be a useful strategy for a game-in-progress!]


Puzzle:

<snip>I think we could have an alternative to a direct mass claim by simply claiming "main game" or "extension", which gives about nothing to scums and allows us to get a better overview of the setup.<snip>

What advantage would there be in doing this today? Or on any day when we no-lynch? The drawback, incidentally, is that it gives the mafia a chance to discuss their claims at night to more effectively coordinate with the better overview of the setup. The only reason I can see would be to confirm a 16:4 setup for the purpose of better determining nightactions, but if there's a 16:4 setup we can massclaim on a later day to achieve a victory almost irrespective of what we have done previously.

So I think it far better to play on assuming an effectively indeterminate setup, in which case the only thing worth discussing today is whether nightchoices should be strategized, and if so, how.

[Puzzle]

Also, the Mafia has no choice but to hope in the presence of extensions (else they've pretty much lost). Although we may confirm it for them by the claim I suggested, it won't change the way they'll play. It may reinsure them but it doesn't give them anymore option.
Actually, after let's say 4-5 persons claim extension, we could stop the process to judge the situation then.

I'm also not a fan of how you present no lynch as the only way forward and particularly, I'd like explanations on the following :"Thirdly, if the optimal play is a no-lynch, there shouldn’t be any bandwagons." :
- Please explain the "if", as you seem dead sure and to be pushing for a quick no-discussion No-Lynch day (that's the impression I'm getting of you right now).
- Please explain how a bit of wagonning hurts the town. Why shouldn't we try to get scums also by "standard" means in parallel.

In any case, the No Lynch wagon is a bit fast to my taste. Could we possibly cool down on it while we decide what to do exactly ?
If the majority wants to go with it, we can still do it after a bit more discussion.

[VisMaior]

So you advocate against the no lynch? I think if we out mafia by discussion we are likely to hit town in the first day, as there is not much to go on. We need a long day 1 to make our chances significantly better than random, so by doing that we will be approximately at the same chances that your first plan gave us. I think it would be better if we agreed on some civs (not players) to advance into the next era for their UUs, while others investigate. This gives us the best chance IMHO in the long run, as we do not sacrifice much of investigative power (especially if we do standard lynching too), and gain a lot the next night.

[Puzzle]

Yes, I'm advocating against No Lynch.

The number of civilization names is limited, even with extensions. Counter-claims are therefore a good weapon, which we shouldn't throw away so easily.
There is also no chance to out a cop or a doc like in standard games.
Even at worst, the loss of a townie wouldn't deny us the possiblity of a chain of inspections in any manner anyway.
Additionally, if the mod gave four townies extension names and gave scums safe claims, I'd rather know it early than too late.

So, yes, I think No Lynch is the wrong way to go.

[dybeck]

The object is to win. I'm sorry if some people don't find it so much fun - but if breaking the game is what it takes, then I'm all for it.

unvote, vote: No Lynch

.

FOS: All those who advocate 'town not winning'

by ignoring the best strategy.

[Astronaut]

And we know for sure that the best strategy is to no lynch, build lots of investigation units and not advance into the next era?

Dammit, I'm always ahead in the technology race when I play Civ...

[VisMaior]

Actually, I think I showed that No Lynch and investigative unit building has approximately the same chance of succesful scumfind as an educated guess lynch, only that the educated guess happens today, and we still can do the investigation thing, effectively doubling our chances to find scum. Is there an error in my flow of toughts? I think not.
Also, some people advancing others investigating might be worth a look into, as I pointed out earlyer.

[Mr. Flay]

The object is to win. I'm sorry if some people don't find it so much fun - but if breaking the game is what it takes, then I'm all for it.

unvote, vote: No Lynch

.

FOS: All those who advocate 'town not winning'

by ignoring the best strategy.

Wow, nice way to try to railroad the town. We've come to no such decision that No Lynch is the way to "break" the game, we don't (or at least I don't!) have any idea what the overall setup is yet. We know there are Japanese, and Barbs, and one other civ, and that's probably it.

VisMajor & Puzzle: I wish I understood half of what you're saying. It seems to make some sense... by the way http://www.kalikokottage.com/civ3/sullla/strat3b.html seems to be helpful to those who haven't played before (if a little hard on the eyes). But I do agree that I don't see why both strats can't be employed while we figure out what's going on... and if it helps, I'm a Main Game civ.

[Puzzle]

Unvote, vote Dybeck

. I was wondering if it was because I was the one targetted by Dybeck's post that I hated it, but Mr Flay confirmed that the post is indeed fairly scummy.

I have a few problems getting on Vismaior's thought lengthwave, as I had understood that he wanted no lynch too, but it seems we in fact agree on having a combination of normal lynch process + circular investigations of the town.
By the way, if some of us didn't make investigation units last night, no need to say so. Just pretend you did for now : no need to indicate to scums who can check them tonight or not.

[VisMaior]

I am generally against No Lynches. I just said, that if we want a good chance on our lynch, we need not to hurry.

[corporateclaw]

I just wanted to check in. I had a 3 day weekend (govt job, woot!), so I was away from a computer. I currently have a large project that must be done, or I die, by wednesday. I wanted to let everyone know that that means I will not be particularly active until wednesday night probably (plus the fact that I'm switching apartments -- insane crazy few days right now!).

I'm not going to random vote or anything because I've not had a chance to read through the thread I don't want to accidentally random vote on somebody that has garnered a lot of votes and push them too close to a lynch without having a reason.

Sorry if this messes it up too much, but I'll be back contributing Wednesday night hopefully -- i suppose unless my boss does kill me.

[EmpTyger]

VisM:

Actually, I think I showed that No Lynch and investigative unit building has approximately the same chance of succesful scumfind as an educated guess lynch, only that the educated guess happens today, and we still can do the investigation thing, effectively doubling our chances to find scum. Is there an error in my flow of toughts? I think not.
Also, some people advancing others investigating might be worth a look into, as I pointed out earlyer.

I think your math in [38] is wrong. 10% chance/night of 1 player finding guilty is more or less correct. (It depends on the specific ratio less any nightkills/blocks.) But the percentage across the entire town is a simple addition: n townspeople gives a (10*n)% chance (assuming everyone is investigate). So with this many townspeople, we have a very good chance of finding guilt.

I’m not convinced a blanket investigation strategy is best, however- it might be better for the town to have some combination of defensive and investigative units. But it’s significantly better than all but the most confident of educated guessing. Which, incidentally, isn’t precluded from being done on a future day if the town’s RNG does happen to be that unlucky. And moreover an educated guessed lynched, if incorrect, would decrease the town’s population. Which would both reduces our effectiveness that night and all future nights.


Flay:

<snip>... and if it helps, I'm a Main Game civ.

I’m pretty sure it doesn’t. In any case, no one else should make a full or partial claim until they or someone else shows that that is prudent.


Puzzle:

<snip>By the way, if some of us didn't make investigation units last night, no need to say so. Just pretend you did for now : no need to indicate to scums who can check them tonight or not.

Er, no one should feign or give *any* indication of what they did last night, regardless of what they did do.

[rajrhcpfreak]

I counted 31 different civs so i dont see how claiming or anything will help much.

http://www.civ3.com/civoftheweek.cfm

[VisMaior]

But the percentage across the entire town is a simple addition

No, that is not true. By that logic only 10 people would be needed for a sure scumcatch, and the rest could rest/advance/build. However many players you put on, you will never reach 100%. Its like throwing a dice 6 times and saying its sure that you got a 6er. (Because every throw has a 1/6 chance of being 6). It simply not how probabilities work. My calculation is the correct way to do it.

[EmpTyger]

VisM:
:oops: Okay, that was embarrassing oversimplication I made. You are right; it’s definitely more complex than simple addition.
But your math is also incorrect.

Let:
I = total innocent
G = total guilty
T = total town (T = I + G)
P = probability of success
s = chance of success for 1 townsperson on a given night
S = chance of success for all townspeople on a given night

The chance s of success for 1 townsperson is given by s = P * G / T
Thus the chance of 1 townsperson failing is 1 - s
The chance that every townsperson fails is (1 - s)^I
Thus the chance S of at least one townsperson succeeding is given by S = 1 - (1 - s)^I

With the 16:4 setup, I = 16, G = 4, T = 20, and letting P = 40%:
s = 8%, S = ~74%

More unfavorable setups are not necessarily worse:
With 1 having died last night and 1 dying tonight (I = 14, T = 18):
s = ~9%; S = ~73%
With 2 dead tonight (I = 13, T = 17):
s = ~9%; S = ~72%

With G = 5 (I = 12):
s = ~12%; S = ~78%
With G = 6 (I = 11):
s = ~14%; S = ~81%

Or let’s consider the absolute worst case, when nobody has an investigative unit tonight so the earliest investigation is Night 3. With 1 dying night 1, and letting 2 die on Nights 2 and 3: (T = 15)
G = 4: (I = 11)
s = ~11%; S = ~71%
G = 5: (I = 10)
s = ~13%; S = ~76%
G = 6: (I = 9)
s = 16%; S = ~79%

And this doesn’t take into account multiple nights of investigation, access to better units increasing P, the chance of multiple successful investigations, the ancillary benefits of the Night 1 nightchoices, or the benefits of successful investigation of innocents.
QED?

The more I look at this, the more a blanket investigative strategy appeals.

[inHimshallibe]

Well, now. I know what a themed game is, thank you very much, and I'm playing in several as we speak.

I'm willing to follow the plan; I'd just

prefer

we play a themed game the old-fashioned way while interacting with the game mechanics.

[VisMaior]

Ugh, you lost me there. Where is

my

error?

[Thok]

VisMaior-You haven't made an error, EmpTyger is just trying to flood you with numbers. His point is that we shouldn't fear the worst case scenario, and the average case scenario seems to be quite beneficial for town.

EmpTyger-I'm not sure whether cops or docs are better to build early-there's something to be said about try to shut down night kills.

[rajrhcpfreak]

too many numbers....

well are you saying that we should all be investgative or we should be spread out?

[Mr. Flay]

I'm not sure how all-investigate can be a failure, assuming some sort of rational basis for the cycle of investigations. BUT, does anyone here have experience with Dethy-variant games (all cops, varying sanity/reliability)? I don't, but I figure that's what we'll be up against...otherwise we're buh-roken like whoa.

[EmpTyger]

VisM:
We each were making [different] errors in combining the chance/townsperson/night to determine the total chance for the town. I’m not completely sure how you got your numbers, so it’s a little hard for me to pinpoint why they’re wrong, but I’m pretty sure it was in tabulating the cumulative probability.


rajrhcpfreak:

<snip>well are you saying that we should all be investgative or we should be spread out?

I’m honestly not yet sure. Instinctively I’m uncomfortable about making the antitowns’ nightchoices easy by publicizing our blanket nightactions. On the other hand, the math seems too good. I’d have to do more math to determine whether there is any advantage (and if so, how much) in diluting the investigatives with protectives.

[VisMaior]

Ok, I see the difference now. you just calculated that 1 townsperson succeeds in investigating, while I calculated that at least 1 scum is catched. Totally different numbers.

[EmpTyger]

Ok, I see the difference now. you just calculated that 1 townsperson succeeds in investigating, while I calculated that at least 1 scum is catched. Totally different numbers.

No, I am calculating that at least 1 scum is caught. Can you tell me what formula you are using to generate your numbers? That might help identify the error.

[inHimshallibe]

Bah and blast to numbers. Just do something.

And tell me what to do.

[VisMaior]

Ok, I got the error. I assumed for some weird, totally stupid reason that investigations only happen if the previous one fails. Obviously that is wrong. EmpTy has the correct numbers. I think the best results could be achieved if we would agree on who is checking who, then it could not happen that a scum gets no investigation at all, increasing our chances. So, who is gonna pick that?