Optimal seer/bodyguard/duke strategy thread

[st.cronin]

In a standard, basic, game, what is the optimal strategy for these roles?

Duke - seems obvious, use it if you are going to be lynched.
Bodyguard - Obviously guard yourself night 1 and 3, guard the seer or duke if they are revealed. On night 2, probably try to guard somebody high on trust lists (if possible), or at least somebody who did not recieve votes on day 1.
Seer - On night 1, it seems like there are two ways to go - scan somebody who recieved votes on day 1, or scan somebody who did not recieve votes on day 1. I'm not sure which is optimal.

Let's have more thoughts.

[Lathum]

seer- scan a wolf

[hoopsguy]

I disagree with guard yourself Night 1 as the bodyguard. It depends on how likely you feel you are to be a night target. If you were in the mix for votes on Day 1, chances are you aren't going to be clipped at night.

I'm also more than willing to not guard someone for a night when they are revealed, with the expectation that I can double back around for them the following day. Doesn't always work - wolves do take the obvious target some of the time - but more often than not it hasn't been a "d'oh!" decision for me to go elsewhere.

Seer - I'll test my instincts on the first couple of nights. Depending on results and game structure, I determine whether I'm looking to build a Circle of Trust or catch a wolf. Mid-game, it can be very powerful to reveal 3-4 villagers and provide clarity - it can be just as valuable as getting a wolf.

Duke - burn it if you strongly disagree with a decision that the group has made. You prove your villager-ness on the spot and introduce a cat/mouse game with the wolves that distracts them from gunning for the bodyguard or seer.

[st.cronin]

I know I have played at least one game where the bodyguard was killed on night 1. I don't ever remember a succesful bodyguard block on night 1.

[Barkeep49]

Originally Posted by st.cronin

I know I have played at least one game where the bodyguard was killed on night 1. I don't ever remember a succesful bodyguard block on night 1.

Statistics say this is not surprising.

[Alan T]

I think a seer is best off trying to guess which people will be around the longest (not likely early votes and not likely early night kills). By picking people whom will be around a bit longer, you might find a lucky UtR wolf, as well as have a day 5-7 reveal with a pretty nice circle of trust.

If you go after likely targets early on you likely end up in the predicament of revealing early for a wolf that would have been a likely vote choice anyhows.

As for the bodyguard, I've had games where I guarded myself on night 1, but that usually is if I dont have a good feel for others to guard. I often try to guess at who the night 1 kill will be and usually its not me. (As I seem to draw alot of early game attention from people). I've gotten lucky twice and blocked on day 1 that approach.

I've never been duke, so don't have much to say about that..

[st.cronin]

I guess what I am actually trying to figure out here is just a strict statistical analysis of best play. I realize game situations are fluid, and often in werewolf the best play is the worst play and vice versa. So given, say, 13 players of equal skill that you are unable to read, and the vote goes:

5 player A
4 player B
2 player C
1 player D
1 player E

Who should the seer scan? Player B? One of player C, D, and E, or somebody else?

That's the sort of question I'm wondering if there's a correct answer or not.

[path12]

I have some comments on this but I'm going to wait until after the current game.

[Passacaglia]

The important part about being the duke is to remember that you're the duke. I think I forgot about it for a few days, then got night-killed fairy well into the game, maybe something like Night 4? So I never got to use it.

[Lathum]

I try and get lynched early as the duke so I can use my power.

After that the wolves leave you alone and look for better targets and the villagers obviously have you cleared. You can play the game without repercusions for the next few days

[st.cronin]

See, looking at this as a bodyguard, and leaving game situations out of it, lets suppose 12 players, 2 wolves.

If you guard yourself, there is a 1/10 chance you will block a night kill. If you guard somebody else, there is a 9/11 * 1/10 chance you will block a night kill (since there is a 2/11 chance you will pick a wolf).

Looking at it as a duke, there is a 1/1 chance that if you unlynch yourself (as Lathum describes) that you are unlynching a villager, and thus a x% chance that you are lynching a wolf. With any other player, there is an x% chance that you are unlynching a wolf, which is obviously not optimal.

Those two are easy to figure out - but, the seer role is obviously more complicated.

[st.cronin]

bump

[Alan T]

I still think I go with my previous thoughts on the subject for seer. To add to it, I do know that as a seer I try to curveball it a bit to try to not be too predictable for the wolves to guess I might be the seer, so usually my votes don't really corespond with what my scans are doing from day to day unless I have something to go on. If I don't know who is or isn't a wolf yet, I try to put my vote in places that might help me learn more about who I feel might be a wolf (ie: trying to vote to push the vote closer, or just go off of my gut).. I won't necessarily then scan that person that night though, depending on what happened with vote movement and such.

Thats just what I do alot of time, but most important for me as a seer is to try to not be predictable in a way where people whom have played a million games with me can get a read that I might be the seer.

[Neon_Chaos]

seer - always scan hoopsguy.

nuff said.

[hoopsguy]

Neon, I feel like seer's nearly always employ optimal strategy if that is the rule.

[Alan T]

Originally Posted by hoopsguy

Neon, I feel like seer's nearly always employ optimal strategy if that is the rule.

I never scan hoops. I just assume that if I am good, he must be evil.

[Narcizo]

[Passacaglia]

Originally Posted by st.cronin

I guess what I am actually trying to figure out here is just a strict statistical analysis of best play. I realize game situations are fluid, and often in werewolf the best play is the worst play and vice versa. So given, say, 13 players of equal skill that you are unable to read, and the vote goes:

5 player A
4 player B
2 player C
1 player D
1 player E

Who should the seer scan? Player B? One of player C, D, and E, or somebody else?

That's the sort of question I'm wondering if there's a correct answer or not.

Player B.

[claphamsa]

There is never a correct answer, if we could say that, then the wolves would just kill whoever your gonna scan every ngiht

[Passacaglia]

Originally Posted by claphamsa

There is never a correct answer, if we could say that, then the wolves would just kill whoever your gonna scan every ngiht

Well, sure. But cronin knows that -- I'm just taking his question and stripping it of its details.

[st.cronin]

Originally Posted by Passacaglia

Player B.

Can you show your work, please.

[jeheinz72]

In that scenario I'd scan whoever I suspected most out of C, D and E.

I'd figure if A gets lynched we get a snippet of info about B (likely the other side from A, or villager-villager), so I'd then look for a new direction to take things.

[path12]

Originally Posted by jeheinz72

In that scenario I'd scan whoever I suspected most out of C, D and E.

I'd figure if A gets lynched we get a snippet of info about B (likely the other side from A, or villager-villager), so I'd then look for a new direction to take things.

I'd likely go this way also, though it sort of depends on what point you're at in the game. Day 1 I might look at B, Day 5 I've already got some opinion on B and as heinz said would get more info from the lynch result on A. So I'd rather get confirmation on one of the other players.

[st.cronin]

My updated thoughts:

Bodyguard, I still think you use it on yourself whenever possible, unless you think you know who the seer is.

Seer, its important to leave behind a trail of information if you get night-killed before revealing. Its also important to scan players who will likely be around at the end of the game. I think in general seers come out too soon if they catch a wolf. I've done this myself, at times.

Duke, I think you primarily use it to save yourself, but I also think you should use it early if you have a hunch.

[Barkeep49]

I think that if you catch a wolf you come out as the seer. Period. The 1-1 is worth it. If you catch a wolf, and die before revealing, it's very hard to discern any sort of breadcrumb trail.

[st.cronin]

Originally Posted by Barkeep49

I think that if you catch a wolf you come out as the seer. Period. The 1-1 is worth it. If you catch a wolf, and die before revealing, it's very hard to discern any sort of breadcrumb trail.

Maybe sometimes, but I don't think that should be some kind of rule of thumb. Lets say you catch a wolf night 1. Vote that player every day from day 2 on, if you get night killed at some point the village should know what to do.

[The Jackal]

Originally Posted by st.cronin

Maybe sometimes, but I don't think that should be some kind of rule of thumb. Lets say you catch a wolf night 1. Vote that player every day from day 2 on, if you get night killed at some point the village should know what to do.

That's the problem though, if you do that the wolf is going to easily pick you out as the seer, or at least they should, so you'll die anyways, without any concrete evidence. And smart wolves can try and work around that.

[The Jackal]

dola

and if you reveal and there's one or more bodyguards, or items, it can make the wolves have to make a really tough decision about the night kill, because they don't want to waste it on someone who is almost certainly protected.

that being said, i don't think there should be a rule of thumb about anything. it's up to the individual player to make the decision within the framework of each particular game.

[st.cronin]

Its really very simple, in my opinion - if the seer is night-killed, you lynch the last player he voted for, and trust any player he has previously voted for. In other words if I am the seer and my votes are Barkeep, Jackal, hoopsguy, and then I'm killed - you should presume BK and Jackal cleared, and lynch hoopsguy.

I agree there are circumstances where a reveal is called for, particularly if there's reason to believe that a bodyguard is still around.

[Alan T]

Originally Posted by st.cronin

Its really very simple, in my opinion - if the seer is night-killed, you lynch the last player he voted for, and trust any player he has previously voted for. In other words if I am the seer and my votes are Barkeep, Jackal, hoopsguy, and then I'm killed - you should presume BK and Jackal cleared, and lynch hoopsguy.

I agree there are circumstances where a reveal is called for, particularly if there's reason to believe that a bodyguard is still around.

If I am ever the seer, people shouldn't assume that when I die. I don't always scan whom I vote for because of reasons I mentioned earlier. Often times the person I distrust the most has no traction for a vote, and i can't get anyone else to vote with me, so I move my vote somewhere that is more meaningful for that day, but still scan the person I wanted to go after originally.

[The Jackal]

When I've been the seer in the past, I don't think I've ever held to a "whoever my vote is on is who I'm scanning" policy. I wonder how many people use this as the norm?

[st.cronin]

The point isn't necessarily that "whoever my vote is on is who I'm scanning." Its that, if I voted The Jackal day 3, get killed night 3, you at least know I was suspicious of The Jackal. And suspicion from the seer is important information. Likewise if I voted for somebody on day 1, then gravitated to somebody else on a different day, it doesn't necessarily prove that I have scanned that person - but it does indicate an elevated level of trust, which, again, is important when coming from the seer. So its not about absolutes - its about probabilities. If you're the seer and you play the way Alan does, not necessarily voting your convictions, then I think you would also play the way Barkeep recommends and reveal early if you find a wolf.

[Barkeep49]

Originally Posted by st.cronin

The point isn't necessarily that "whoever my vote is on is who I'm scanning." Its that, if I voted The Jackal day 3, get killed night 3, you at least know I was suspicious of The Jackal. And suspicion from the seer is important information. Likewise if I voted for somebody on day 1, then gravitated to somebody else on a different day, it doesn't necessarily prove that I have scanned that person - but it does indicate an elevated level of trust, which, again, is important when coming from the seer. So its not about absolutes - its about probabilities. If you're the seer and you play the way Alan does, not necessarily voting your convictions, then I think you would also play the way Barkeep recommends and reveal early if you find a wolf.

I think suspicious voting patterns, as you seem to encourage, is a great way to draw villager suspicion upon yourself.

[st.cronin]

Originally Posted by Barkeep49

I think suspicious voting patterns, as you seem to encourage, is a great way to draw villager suspicion upon yourself.

I concede that different situations demand flexibility, but I guess I don't get what can possibly be suspicious about voting patterns based on seer scans, which is what the seer will be doing regardless. I don't ever recall a seer being lynched or nearly lynched because of his suspicious voting record, although I'm sure its happened.

On the other hand, the village is nearly always suspicious of me no matter what my role/allegiance, and no matter how I play, so I usually look at my votes as being very important statements, whereas other players often use their votes to avoid standing out.

[jeheinz72]

When I'm the seer I definitely don't lay such a trail to follow.

I generally try to play dumb-as-hell until I get a hit on a wolf so the baddies avoid me. Then when I get a hint, I"m more apt to drop a comment on the person than a vote, or reveal if the stage of the game is right.

[Passacaglia]

Originally Posted by The Jackal

That's the problem though, if you do that the wolf is going to easily pick you out as the seer, or at least they should, so you'll die anyways, without any concrete evidence. And smart wolves can try and work around that.

I don't agree with that. If I'm a wolf, and someone votes for me every day, I think hard about whether it's worth it to kill them, and put suspicion on me -- whether that person is the seer or not.

[Passacaglia]

Originally Posted by st.cronin

Its really very simple, in my opinion - if the seer is night-killed, you lynch the last player he voted for, and trust any player he has previously voted for. In other words if I am the seer and my votes are Barkeep, Jackal, hoopsguy, and then I'm killed - you should presume BK and Jackal cleared, and lynch hoopsguy.

I agree there are circumstances where a reveal is called for, particularly if there's reason to believe that a bodyguard is still around.

That just doesn't jive with me -- even though I've used this argument when you were the seer before. If your vote isn't made with much fervor, I don't think people are going to pick up on it as being a seer scan (unless they want it to be, as in the aforementioned argument).

[hoopsguy]

Originally Posted by st.cronin

See, looking at this as a bodyguard, and leaving game situations out of it, lets suppose 12 players, 2 wolves.

If you guard yourself, there is a 1/10 chance you will block a night kill. If you guard somebody else, there is a 9/11 * 1/10 chance you will block a night kill (since there is a 2/11 chance you will pick a wolf).

Looking at it as a duke, there is a 1/1 chance that if you unlynch yourself (as Lathum describes) that you are unlynching a villager, and thus a x% chance that you are lynching a wolf. With any other player, there is an x% chance that you are unlynching a wolf, which is obviously not optimal.

Those two are easy to figure out - but, the seer role is obviously more complicated.

I agree with the math you use for the bodyguard, but I don't value my role all that highly early in the game. If I'm not a likely target to be night-killed (and I'm usually not all that high, based on generating suspicion early in games) then I would rather try to protect someone I think may be the seer or someone I think is a more likely kill. I understand that I'm giving up some percentage value - as you illustrate - but my primary goal is to get a block and my secondary goal is to protect important roles.

Mid-game, when the player list is down to 7-10, then I'm much more inclined to protect myself at every opportunity unless given a compelling reason to go elsewhere (revealed seer, for example). At that stage I want to see if I can last until the end of the game and dramatically increase my percentages of getting a good block each and every following evening. Personally, I now think that I'm more likely to be killed before the final 3-5 players as wolves often want to assume control of the conversation at this stage in the game.

End-game: if I'm still alive at this point then I'm a huge threat to the wolves. I've usually revealed my role by this point. It is just a game of chicken at this point and I have not really developed any hard and fast rules on how to play it.


I recognize I'm not advocating the "mathematically correct" way to play the role, but I do think that I'm maximizing the way to play the role for me individually, based upon style of play and perceived reputation in our community. I would probably go more by the numbers if I was to play werewolf under a different handle or on another site where I had no reputation.

[st.cronin]

I'm starting to think that, as a seer, what you should do is rate the players in order of MOST suspicious to LEAST suspicious - and then scan the players right in the middle of that spectrum. The players who are most suspicious are likely to be lynched, the players who are least suspicious are likely to be night-killed. This is probably what Alan was getting at when he said he tries to figure out who is most likely to be around at end-game.

[path12]

Originally Posted by hoopsguy

I agree with the math you use for the bodyguard, but I don't value my role all that highly early in the game. If I'm not a likely target to be night-killed (and I'm usually not all that high, based on generating suspicion early in games) then I would rather try to protect someone I think may be the seer or someone I think is a more likely kill. I understand that I'm giving up some percentage value - as you illustrate - but my primary goal is to get a block and my secondary goal is to protect important roles.

Not to sidetrack too much, but I think that you not getting night killed early is not so much based on generating suspicion, but more a sense that you're likely to be guarded or a seer target early and more of a risky night kill for that reason.

I know that's always the dilemma I face when a wolf in a game you're a villager in.

[Narcizo]

Originally Posted by hoopsguy

I'm also more than willing to not guard someone for a night when they are revealed, with the expectation that I can double back around for them the following day. Doesn't always work - wolves do take the obvious target some of the time - but more often than not it hasn't been a "d'oh!" decision for me to go elsewhere.

Seems like a big risk. I guess it depends on the game situation. If you guard the seer then you're guaranteeing another scan on a player. If you wait a day you could be guaranteeing two scans but you might lose anything. It certainly won't work if the bad guys have reason to believe that there isn't a bodyguard.

[Narcizo]

I was was also wondering about the thinking of hinting at being a seer if you're a bodyguard. I guess generally speaking it's too risky as the wolves might wait a night and then kill you, but has anyone had any success doing that?

Also, if you're someone like Alan, who everyone expects to come up with lots of good theories how should you play it if you have a good role. Seems like it's a tough spot to be in, as you don't want to attract more attention but you don't want to be lynched for not playing your usual game (or make the wolves suspicious by doing the same). But you might not want to, inadvertently, scare a wolf by going after them.

[Alan T]

Originally Posted by Narcizo

I was was also wondering about the thinking of hinting at being a seer if you're a bodyguard. I guess generally speaking it's too risky as the wolves might wait a night and then kill you, but has anyone had any success doing that?

Also, if you're someone like Alan, who everyone expects to come up with lots of good theories how should you play it if you have a good role. Seems like it's a tough spot to be in, as you don't want to attract more attention but you don't want to be lynched for not playing your usual game (or make the wolves suspicious by doing the same). But you might not want to, inadvertently, scare a wolf by going after them.

I had a great game as Bodyguard in Hoops' Marvel game where I seemingly guessed the wolves steps two nights in a row with blocks until they killed me (I couldn't guard myself).. but the other two games I was a bodyguard, I don't remember the circumstances, but both times I think i did what Hoops suggested he would do and try to outsmart the wolves only to be burned bad.

As a seer, I know that I've been killed like day 1 or 2 several times as a seer, and still to this day have no idea what I did in those games to tip off being the seer, or if they just got lucky! So I haven't really ever been too good at being a seer I don't think.

[hoopsguy]

Originally Posted by Narcizo

Seems like a big risk. I guess it depends on the game situation. If you guard the seer then you're guaranteeing another scan on a player. If you wait a day you could be guaranteeing two scans but you might lose anything. It certainly won't work if the bad guys have reason to believe that there isn't a bodyguard.

Pretty much every game has a seer and a bodyguard. So unless there is a "no roles revealed on death" I think the wolves always assume they are working to avoid a bodyguard.

It is absolutely a big risk. But if the wolves know where you are going to be then they are free to kill anyone else. So I don't think you can absolutely guard the "right" target every time ... I think you have to introduce enough variance to give yourself a chance to fulfill your primary objective (get the block) but not too much that they can go 2nd level thinking on you ("he won't guard the obvious guy, so we'll kill the obvious guy").

[Narcizo]

Originally Posted by Alan T

As a seer, I know that I've been killed like day 1 or 2 several times as a seer, and still to this day have no idea what I did in those games to tip off being the seer, or if they just got lucky! So I haven't really ever been too good at being a seer I don't think.

I wouldn't imagine you have to tip them off to you being a seer, you're a likely early kill target anyway I would have thought. (based on games I've been involved with you - ie you've won 2 out of 3 of them for the village).

Maybe you should be going fairly hard after someone you don't think is a wolf, so the wolfies will keep you around barking (geddit!) up the wrong tree for a while. Not too hard or they'll kill you to make the guy you're going after look suspicious.

[Narcizo]

Originally Posted by hoopsguy

Pretty much every game has a seer and a bodyguard. So unless there is a "no roles revealed on death" I think the wolves always assume they are working to avoid a bodyguard.

Last two games have had multiple bodyguard roles (I think) whatever initially revealing that there are multiple roles. That's the sort of thing I was refering to.

I believe "no role reveal" games have fallen out of favour with pretty much everyone. Understandably as it removes voting pattern analysis.

[Alan T]

Originally Posted by Narcizo

I wouldn't imagine you have to tip them off to you being a seer, you're a likely early kill target anyway I would have thought. (based on games I've been involved with you - ie you've won 2 out of 3 of them for the village).

Maybe you should be going fairly hard after someone you don't think is a wolf, so the wolfies will keep you around barking (geddit!) up the wrong tree for a while. Not too hard or they'll kill you to make the guy you're going after look suspicious.

I used to be an early kill choice and early vote choice often, but for the past year or more I haven't really been. I think its because occasionally I have the good game that you reference, but just as often I seem to be way off track on who I think is bad, and it is useful for them to keep me in there to drive everyone the wrong direction.

[Passacaglia]

I think the thing about us hinting at being the seer is that since we were both bodyguards, Narcizo and I could guard each other. Also, if one of us revealed as the seer, then someone like PB came out as the seer, we could have alternated between the two, and kept two people alive every day. Uh, as long as one of us didn't get blown up. Oh, and as long as the other one of us doesn't get converted! Would have been interesting if Narcizo had been converted while we could still PM each other.

[Narcizo]

Yeah, I was thinking that. Although you probably wouldn't have been around much longer after that, depending on when you last protected yourself.