Dominion Strategy Forum

Please login or register.

Login with username, password and session length

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - SpaceAnemone

Filter to certain boards:

Pages: [1] 2 3 ... 189
@Space, Why are you assuming scum!Dylan even had a question?

I wasn't, but if scum!Dylan was making up the existence of a question, why would he offer to use that to cop someone for us? Robz's name was one of the first to come up, and we know he's town, so why wouldn't lying scum do something that's easier for scum to steer for their own scummy uses than asking for town input? If town recommends using the cop shot on someone who's town, Dylan either has to put himself in a 1-v-1, or to create an almost-IC. Instead, scum!Dylan could just have faked the result of a cop shot on scum (if that was his aim?), rather than faked asking for advice on a cop shot.

The only explanations I can think of feel really weak, like maybe scum wanted to get town's reads on everyone, or maybe there really was a one-bit question, but scum were the ones who were given it and just wanted to use the thread's creativity to work out the optimal use. Otherwise, I could do with some more-likely-sounding suggestions for what Dylan was up to with the question if he's really scum.

I don't know. But that is easily answered if you're scum. That question is hard to answer, and I've considered that my power might've been tampered with because of it. I just don't see town winning if scum can force a direct conflict between 2 town like this, and it makes my power borderline anti-town. But my paranoia's somewhat there.

It's not very easy to unpick what you're trying to say here. My question was about why scum!Dylan shares the existence of this question and lets town direct it when it's likely to lead very quickly to one of two town-beneficial events. What "conflict between two towns" are you referring to, and what would scum!Dylan's plan have had to look like to bring that about without risking being pro-town?

Why would scum kill a third-party? ADK being alive is their best chances of winning this game.

Ah, so please explain in detail why we're shouldn't all be voting ADK now.

So we can kill scum instead.

So far we've done a great job of going after townies rather than scum, so if ADK being alive really is scum's best chance of winning, over pushing to get townies exiled, then I think that's a great argument for exiling ADK instead of letting scummy people or wrong townies take down another one of our own.

It's a very good question, and one that pertains to ADK's alignment, so I won't answer in detail. But scum not wanting to shoot a third-party doesn't mean we want to exile them. Town wants to kill scum. Scum wants to kill town. Neither wants to kill third-parties.

Well, I think I might want to kill a third party, and I'm town.

Why would scum kill a third-party? ADK being alive is their best chances of winning this game.

Ah, so please explain in detail why we're shouldn't all be voting ADK now.

* why we shouldn't all be voting ADK, I mean. I was going to say "why we're not all voting ADK", but that wasn't as clear.

Why would scum kill a third-party? ADK being alive is their best chances of winning this game.

Ah, so please explain in detail why we're shouldn't all be voting ADK now.

I think Iím ready to vote for Dylan.

But I donít want to rush folks if we arenít ready for that.

I don't want to rush it.

If Dylan is scum, why did he offer up a one-bit question that he was proposing using as a cop shot that would either have given town an IC, or set Dylan up for a 1-v-1 with whomever he accused, that was bound to get him exiled today anyway?

There's 9 players alive: if both me and ADK are third-party, chances are town has already lost, and we've already won. At this point, you're going to have to trust us when we say we want town to win.

OTOH, scum isn't taking ADK out, but is clearly able to perform NKs, so how can we be sure ADK really has town's interests at heart, if keeping you and ADK alive is so townie?

Scum!Robz would never stop playing.  He'd sub out if he got too busy. Your assertion that he claimed scum didn't make sense.

Did MiX ever explain to anyone else's satisfaction what Robz's alleged admission of scumminess was?

No, by definition.

I asked 3 times what Robz had targeted and what he saw. He never answered. That felt like he was taking his lurkiness to extreme levels, to the point where if he was scum, he didn't need to talk anymore. That's basically it. I was also annoyed by Robz not having participated basically at all from D3 forward.

That was all? This is why you need to actually show your logic in detail in-thread :-( You being annoyed or frustrated at Robz for ignoring or missing a question from you doesn't make him scum in the slightest, and I think your push against him at the EoD was not the best use of town's information at that point.

Why are we not exiling Jack?

I actually agree that Jack isn't the best target here. Scum choosing to kill within yesterday's testing set is unexpected to me, and probably only happens because they're happy enough with the idea that town is likely to exile Jack.

I have next to no confidence in MiX's claims, but I do actually think that exiling Dylan isn't terrible for town, simply his flip tells us whether to trust the answer to his question, and that's the only piece of logic-based scum-hunting info we have left that's not influenced by claims coming from MiX and ADK.

As for MiX and ADK, I think there's a very strong chance they're working together towards their own goals, but I think it's less likely they're cylon-aligned than purely a third party, simply because MiX has been leaking some kind of third-party knowledge all game. However, I do know for sure that unless scum were deliberately able to influence the cylon test then MiX did as he said he would do, and set his side to cylon, since that's the only way we'd see a cylon result when I set my half to read human.

Scum!Robz would never stop playing.  He'd sub out if he got too busy. Your assertion that he claimed scum didn't make sense.

Did MiX ever explain to anyone else's satisfaction what Robz's alleged admission of scumminess was?

Last night ash and Space had the qt.

Yep, I can confirm that part.

Today is busy, so no more content from me till after 3pm FT at least.

Dylan is scum

Vote: Dylan

Let's go let's go everyone needs to vote for dylan right now before scum quickhammers and wins!

Last game day you said Robz was definitely scum.

I'm not believing another MiX statement that isn't backed up with sound logic described in-thread.

PPE 2, but it's work time for me so I'll have to catch up at lunch.

The first person to get on the thread should do the test, most likely.

I think EFHW specifically has to do the test, unless we're all so certain she's a cylon that we move onto the next-most-suspicious person?

As evidence, LL was town, so was telling the truth about being able to detect cylons. He knew EFHW targeted him D1, so previously I thought either she must have been a cylon, or LL was lying. With LL's flip, we at least need to hear from EFHW about why she thinks that is. We've already seen on human-aligned cylon in the form of Boomer, so I'm pretty sceptical about the existence of a second, as opposed to actual cylon-aligned cylons.

So here is the result from my claim, phrased carefully to communicate exactly what I asked and the answer without quoting anything.

There is an odd number of cylon-aligned players in {MiX, Space, LaLight, Jack, Didds}.

Awesome! Thank you for going through with that test :-) If we assume this is true, and that there are 2-3 cylons overall, that almost certainly means exactly 1 from the test group, and then 1-2 from {ADK, Ash, EFHW, Scola, Robz}.

I would propose that we start with a Jack exile, as I suggested earlier. If Jack flips cylon-aligned, that gives us more trust in LL's results, which I think suggest that EFHW is a cylon, though I  would really like some more confirmation on that before proposing that she should be under scrutiny tomorrow. If Jack flips not-cylon-aligned, I guess LL would be my next pick, though of course there's also the obvious choice of flipping Dylan just to be sure.

On the off-chance there's an early rush for deadline since a lot of players are in US timezones, I'll drop a vote now:
Vote: Jack

Also, MiX, I intend to calibrate to "human" tonight, so would you like to do the complement to that so we can finally get this test working? If we have a possible third controller, maybe the best they could do would be to flip a coin in case we're messed up by RBs or worse.


does anyone mind if I go for it myself, just so we don't completely waste the opportunity to use it and see what happens?

Yes, I mind, the person in charge of testing shouldn't be the one tested.

Why not? Once we've set it up, it's a public resource and anyone can submit. I think if I refuse to be tested, I'd be setting a double standard: maybe the test will suggest I'm a cylon and have suspicion levels raised, but everyone who's tested faces that same risk, so there's nothing special about being the one who calibrates it.

My "if nobody minds" was more geared towards finding out if anyone (maybe Ash?) had some desire to take the test instead, if we're establishing that EFHW is refusing. I don't think I have a lot of respect for a suggestion that I should be banned from taking the test on principle if nobody else wants to do so and I do.

Also, you should probably say what you picked before we test, I don't remember if you did.

Oh, I thought I'd said that I'll reveal that only after we run the test. Given that we don't expect the test result to be useful except in revealing if it's been tampered with, and since town has to take my word for it ether way, I think there's no more information given out if I say so before or after, is there? If that's the case, I'd prefer to be the one interpreting whether a possible difference between what I say and what we observe is due to me lying or something else messing with the power :-)

Honestly it doesn't matter, but I think it would be better to not do it at all than do it in a way that is questionable.

It was questionable both nights so far, in that we have never hit the one-human-one-cylon setting, unless you lied about what you set it to. I think if there's a risk of RBs or similar out there who could mess with our ability, it will always be a little bit questionable. I just want to see now whether we see quirks that might make us more or less trusting of future results.

PPE 1: exciting result!

Regarding testing me, I don't see the point today, but would becwilling tomorrow assuming my continued survival.

The point today would be to see whether the test comes up as I set it, or if something else is going on. It's not super-useful information on its own, like knowing if there might be a third controller out there, or maybe some kind of scum interference messing with the results. There really doesn't seem to be any down-side to using the test: there's no indication that it's shot-limited by anything other than the survival of the people who're setting it up, and both people who've been tested so far haven't reported any ill effects.

I'm kind of puzzled by EFHW's unwillingness to go for it, hence feeling like it's something to push a little. If nobody else wants to push that, though, does anyone mind if I go for it myself, just so we don't completely waste the opportunity to use it and see what happens?

Did that simulation include that there's 3 scum? I don't think I need to prove that:

If there's 3 scum in 10 people, then in 2 5 people groups, scum is in a 0/3, 1/2, 2/1 or 3/0 split, with equal probability of being 1/2 as 2/1, and 0/3 as 3/0. So 0/3 + 1/2 odds are equal to 2/1 + 3/0 odds, which means each of them is 50%. Note that the second sum is the "at least 2 scum in the first group" question.

We are definitely talking over each other at this point, so I don't really see us agreeing on this before the day's over. Dylan should just ask whatever he feels like.

Aha, sorry, yes, the numbers I pasted in were for the group of 4, because I'd been thinking on the side about the point that including Dylan in the 5 isn't necessary. This is why I shouldn't start something technical then go back to work and forget my train of thought :-/

To summarise using the conventions you used above: for the "2 or more" version:
"no" -> 0/3 or 1/2
"yes" -> 2/1 or 3/0
-- this relies on there being exactly 3 scum, because it uses a cut-off between 1 and 2 to cut that particular distribution in two.

And for the "is the number even" version:
"no" -> 1/2 or 3/0
"yes" -> 0/3 or 2/1
-- this will work just the same for other numbers of scum, and other sizes of group.
-- it also gives an outcome where within the "no" or "yes" bucket there are two more-different possibilities, and I think that makes it easier for us to work out which case it really is, based on where the scummy people are, e.g. after exiling one from each set and seeing who scum decides to NK.

I thought Ashersky's proposed uses of the question were all fair, though I think they have similar or equivalent weaknesses to the other things we've considered. A straight-up cop shot is pretty much where Dylan opened up the discussion. I totally agree that "baddies in a subgroup" isn't ideal if you don't trust whoever's picking the "baddies", but that was why I was trying to pitch my idea as a consultation with the thread, because talking about who we trust enough to want to put them into a goodies/baddies pot is precisely what we want people to be doing for fruitful gameplay anyway. And of course, if we don't trust Dylan then the whole thing is moot.

As for the game-mechanics ones, I think "exactly four alignments" is the kind of thing faust could swing on a technicality, depending on whether MiX's claimed town-but-survivor is different from my straight-up town role. The "does a town-aligned cylon exist" might be complicated by the fact that it seems likely that Swowl-Boomer was exactly a town-aligned cylon, so does Swowl still exist for the purpose of the question or not? "Is a town-aligned cylon still alive at this present moment in the game" might be a better wording. However, now the mechanism for operating the cylon testing is out in the open, I suspect it's even less useful for us to know whether that's the case, because if all it takes is killing one of me or MiX, then if the cylon-aligned faction is too worried about it, they'll just get one of us killed.

@MiX, I feel like there's very little chance that we're talking about exactly the same thing here without having incompatible hidden assumptions, because I strongly disagree with almost everything you say.

By the laws of deepwolfing, a "townie-reading group" always has 1 scum, and exactly 1 scum.

Does "deepwolfing", mean that one member of a team will always try to behave in as townie a manner as possible? If so, then I think it's probably true on average, but it can't be a "law", because if we assume that as town, then scum could get an edge simply by refusing to have any super-townie members, thus being able to count on town exiling one of their own to fill the quota.

I don't see why using reads to make questions 50/50 is better than using the law of probabilities to make 50/50 questions. I hope you agree with the following statement: "If there are exactly 3 cylon-aligned players, a question that asks whether there are at least 2 cylon-aligned players in a group of 5, assuming Dylan is town, is a 50/50 question". This seems much more exact than your argument about the 50/50-ness of your question.

First up: where is all your confidence in exactly three scum coming from? Do you have some extra game information that I missed?

Second up, let's assume a completely random group of 5 (excluding Dylan), in which each player has a 3/10 chance of being scum. By a rough simulation, I get the following histogram over the number of scums in that random set.:
   0 scums: 16.7%
   1 scums: 50.1%
   2 scums: 29.9%
   3 scums: 3.3%

So if you do "at least two", it's a 33-v-67 question, whereas "odd or even" is 53-v-47. What's more, once we have an odd-or-even, we're left in a much easier game state, because telling the difference between 1 and 3 is easier than telling the difference between 2 and 3, and I think normal game-play (like what we're going to have to do anyway, since we can't actually trust that Dylan tells us the truth) helps more in determining where we are. It's that ability alternate between states giving a "yes" and those giving a "no" that make me think that something parity-based is going to be more helpful to us than anything with an inequality is.

Of course, the above is illustrative only, because of I don't want to assume exactly 3 scums, and in real life we can use reads to at least partially sort out the list to make our expected histogram look a bit different. In particular, my initial proposal was meant to have 5 people such that ~50% of the probability is on 1, most of the rest is on zero, and a smaller portion is on 2 (or more).

For example, I'm basically 100% certain there's 1 scum, and only 1 scum, in the target group. That's not really valuable information.

If you can be that certain for any target group proposed, then congrats, we can use you as a perfect scum detector and just win, if you're town.

I started a string of responses in different tabs earlier today, but I've been too busy to finish any of them yet. Here's one that's more or less ready to go:

I realize that we're close to deadline and I'm complicating things even more, but:

Based on my win condition, I have reason to believe that it's very likely that there's at least one scum who is a civilian, and probably not more than one. Right now robz is my best guess for tha, but it's also another subgrouping to consider for dylan's question

Did we lose track of this bit of information?

Oh, yes, a little! Thanks for the reminder :-) I do think it's covered okay in the groupings I suggested, but on the other hand, I think MiX (and implicitly Dylan) being in Testing and Ash and Awaclus being in Remainder more or less works.

For reference, this is what I've got in terms of people's claims, including my assumptions about military/non-military and some testing results, so please let me know if anything is wrong or if it's missing any key pieces:
1. Jack Rudd -- Saul Tigh, military
2. SpaceAnemone -- Gaeta, military
3. WestCoastDidds -- Starbuck, in pilots QT, military
4. MiX -- Baltar, civilian
5. EFHW -- Duala, possibly cylon species through LL's D1 result, military
6. scolapasta -- Bill Adama, military
7. LaLight -- Helo, in pilots QT, military
8. Dylan32 -- Zarek, civilian
9. A Drowned Kernel -- Apollo, tested Human d3, in pilots QT, military
10. Swowl -- Boomer, human-aligned, in pilots QT, military, died n3
11. Robz888 -- Cally, civilian
12. ashersky -- Chief, civilian, tested Cylon d2
13. Awaclus -- Billy, human-aligned, civilian, died n2
14. mathdude -- Roslin, human-aligned, civilian, died mid-d1

As Ashersky has said all along, the civilian-vs-military distinction for the deck crew members (Cally and the Chief) may be a blurry line.

I don't think either of the tests was yet trustworthy:
Mix claimed at #1264 to have picked "cylon" on N1 and "human" on N2.
D2 test: Only "cylon" from Mix? -> Fits with Ash getting cylon result
D3 test: 2*human ->  fits with ADK getting human result
D4 test: Only my setting?

My point is it's completely pointless to include me anywhere in the analysis.

Yeah, I agree. It helped my thinking to assume you were townie and place you  accordingly while reasoning it out. Who's your most-townie guess from the remainder set?

I'll probably wait until tomorrow afternoon to actually ask my question and I'll post here before I do it.

Deadline is about 36 hours out, and while it's after 9pm FT now, there's zero chance of me being awake this late tomorrow: today I was so exhausted I had an after-work nap, whereas tomorrow I'm more or less non-stop busy all day already.

I think MiX, Jack and I are all on FT+5, and Ash is something like FT+15, so as we approach deadline it's worth considering availability of people to react to the result and vote accordingly enough to get an actual exile through.

In what way shape or form is asking if there's an even amount of scum in a 4 player group (not counting Dylan) a 50/50 chance? Odd is much more likely.

What I'm going for is P(n==0)+P(n==2) ~= P(n==1)+P(n==3), i.e. the chance of having zero or two scum is more or less the same as having 1 or 3 scum. You can even add in more scum if you want to be thorough about covering all the game combinations. So mostly it's asking whether, in a townie-reading group, we have 0 or 1, with a small chance of 2. Since 0 counts as even, then 0 ought to be almost as likely as 1. If we really do have zero, our whole test group is kind of IC'd if Dylan ever does flip green, unless we start suspecting later that we're in the much-less-likely 2-scum scenario, but I think the natural gameplay will take care of that, and it's a very precarious position for scum to play in.

For one possible difference between the two variants of the question ("even/odd" vs "at least 2 in X"), consider a case where there are currently only two cylon-aligned players (e.g. because Awaclus didn't join them maybe?), and we have one in the testing set and one in the remainder. If we do a parity-based test, we at least realise there's only one in the test set (or an exceedingly unlikely three), whereas from the "at least 2" test we're left with knowing there's 0-1 on the test side and any number at all on the remainder. I don't think this is so unlikely a scenario, though I do agree that I think three scum overall is more likely.

@Dylan, I'm really quite confident that parity (i.e. the odd/even question) is more informative than any "at least X" test, but I'm also feeling that I'm unlikely to convince MiX at this time of night. Honestly, I can just think of too many maths puzzles where parity is a stronger way of doing things.

If we want to split the player group in two a different way, one alternative I thought of was to split us into a slightly-scummier five and a townier six, then to ask whether there are strictly more cylon-aligned people in the bigger group than the smaller one, but I'm not sure whether that's as good.

One notable change to the groups proposed though - I'm not going to include myself in the testing group, because:
A) I'm town and won't affect parity of the results, so me being in the group is pointless.
B) If I was scum, then you shouldn't trust the results at all anyway, so me being in the group is pointless.

So Space, with that knowledge, who would you move from the remainder to the test group in my place?

I was deliberate in including you with the "maybe-IC" testing group, because we absolutely can't use the test information if we don't think you're townie, so I'd prefer you to stay there, to balance numbers a bit.

I'm not sure whether or not we ought to move anyone else up in your place if you did decide to take yourself out, though... we want to keep the chance of one cylon-aligned player to be more or less the same as the chance of no cylon-aligned players (plus the chance of having two of them), and right now with that set I'm more or less comfortable that I couldn't tell which way the coin flip would go.

Also, I'm getting really tired and should already be asleep, so I trust two-hours-ago me rather more than current-me right now.

I am in perfect agreement with the group, except the question should be "are there at least 2 cylon-aligned players in the group". This way we get to exile in the group that has the most amount of scum right off the bat, and after a scum flip we have 2 groups, each of them have 1 scum.

I think the only difference between what I said and what Space said is that instead of doing a 0/1 cylons group and a 2/3 cylon group, Space does a 0/2 cylon group and a 1/3 cylon group. I'm not sure how this added complexity helps us. I'd wager that knowing where the majority of scum are is enough of an upside to do my suggestion over Space's suggestion.

Wait a second: Space, shouldn't the group with Dylan be the bigger one?

Are you assuming exactly three cylon-aligned players to make "at least two" a useful question? While I think that's likely, I don't like just assuming it as a starting point for anything.

I definitely don't agree that these are the right groups if that's the style of question you want to ask, because I think the expected number of cylon players in the testing group is far below 2, and my entire point was that we get the best information when we ask a question where we're exactly 50/50 on whether the answer will be yes or no. I think what you're suggesting is a strictly worse use of information, from an information-theoretic standpoint. Either you're arguing that optimal use of the question is too hard for you/town to reason with so you don't want to be maximally townie, or you're arguing against wikipedia saying that setting both groups to p=0.5 is the best we can do for info on a one-bit test.

(Also, you're incorrect about your statement that after a scum flip, we have one scum in each pool, because your scenario doesn't distinguish between there being 2 scum in the test group or all three, whereas mine is explicitly set up to cope with telling differences like that, which will be useful for town later, and even more useful if we also add in the exactly-three-cylon-aligned-players constraint that you're silently assuming, since it would let us know that an odd number in the testing group means an even number in the remainder group and vice versa).

Oh, whaaaaat, you get a power that is unknown and you do nothing with it? Awwww.

I barely had time to engage with the game, and it wasn't at all easy to keep up with, so I stand by not touching the questionably-useful power that would make me hated the next day.

There's actually no roles that care about human or cylon (I guess LL's, but theirs also detects), so this power seems a bit useless...I guess we're supposed to exile cylons?

I think we expect a decent correlation between who's a cylon and who's cylon-aligned in this first game. In particular, I think the main thing that stops this power being more useful is the risk of human-aligned cylons, and I think we've already seen one of those flip because of Boomer. It's much less common to have cylon-aligned humans in the show, but Baltar and Helo are two that I could imagine in that boat in some scenarios. Let's just test people other than Baltar and Helo for cylon-ness first.

Okay, in the interests of time, here's exactly how how my proposed version of Dylan's yes-no test might work. This is my current best guess of what gets the strongest info for town, but it's quite possible I've missed things that the rest of you can fill in, so consider this a consultation exercise if you like :-)

Testing group: MiX, me, Dylan, Jack, Didds

Remainder: ADK, Ash, EFHW, Scola, LL, Robz

We would ask whether there's an odd number of cylon-aligned players (making explict that we don't count zero as odd) in the testing group.

Some of my reasoning for this partitioning of the data is as follows:
  • MiX and ADK should be on opposite sides, since one of them being cylon strongly implies that the other is, and having a confusing zero-or-two scenario isn't what we want. I'm putting ADK on the outside since chances are they're third-party anyway.
  • Including one person who town doesn't mind voting for in the testing group gives us a strong place to start if we do get an odd response, and later on it may also be useful in we get an even response now but want to be sure it's not two scum instead of zero. That's why I've put Jack in the test set.
  • I've put three of the people from the Jack wagon into the testing set, and the other three in the remainder. This way if we do decide to trust both this info and MiX's report of at least one scum on the Jack wagon, then the group is split up nicely for that reasoning too.
  • I've put EFHW and LL in the remainder half because there's some odd stuff that doesn't quite gel about their claims.
  • I've put Robz in the opposite group from Jack because there's a reasonable level of support for exiling him already, and if we get a zero result then at least for today I think we assume zero in the test set and exile from that remainder set.
  • I actually feel townier on Scola than on Didds, but I realise I'm in the minority there, so I've put Didds inside the group and Scola outside it.
Remember, it's not like we're able to play as if this is guaranteed-true advice without Dylan flipping green at a later point in the game anyway. So what I'm trying to do here is just balance things up so we expect 0-1 scum in the "testing group". If we get an "odd" response, Jack is probably the first to be exiled. If we get "even", we assume zero scum in the testing group for today, and exile from the remainder group, in which case it currently looks like Robz gets the popular vote, though I guess there's still plenty of room for opinions to shift, and I think there are more people in the outside group that others are willing to vote for anyway.

So, who's in agreement, and are there any name swaps that make sense? If anyone disagrees outright and argues for using Dylan's shot more like a cop on a single player, are you really sure that that would give town more of an advantage than this?

Pages: [1] 2 3 ... 189

Page created in 2.721 seconds with 19 queries.