EPTThe TL [N][M] - Page 4
| Forum Index > TL Mafia |
|
DarthPunk
Australia10857 Posts
| ||
|
Aquanim
Australia2849 Posts
On January 04 2014 17:13 DarthPunk wrote: I honestly am hoping Blazinghand will make a 13 player version of his GSL setup. What exactly are the aspects of the GSL setup which appeal to you, out of curiosity? I don't disagree with you by any means, by the way. The most obvious unique feature IMO is that scum get very little information from the roles they get; if they get a Roleblocker town could have two power roles, or none at all. I'm not entirely sure if that's practical to replicate in a 13p game. Even in the 9p game I think there's a fair bit of difference in balance between a pcop+doctor and 2 namedVT setups (though I wouldn't venture to guess which is more balanced) | ||
|
wherebugsgo
Japan10647 Posts
Let's look a bit at the 3 roles you provided. I briefly had a convo with Foolishness the other day about something similar, in which I asked him what he would rather have: 1 shot cop 1 shot vigi We both agreed that in almost all situations we'd rather have the vigi. Now, what about a full fledged cop and a 1 shot vigi? Now it's closer, and one could potentially make arguments for a vigi (particularly in the hands of a better player) but I think we both agreed again that we'd prefer a cop. One thing I didn't consider, is the following: which would you rather prefer? Parity cop 1 shot vigi I think this one comes down more to personal preference. I think each player would have his or her own preference here, but it's much more even I think than the previous two situations. Similarly there are things like 2 shot cops, but that brings in a host of other issues. When I was tweaking C9++ I thought about this a lot so that the letter rolls would give balanced outcomes, but it's quite a hard thing to think about in the end. Let's look at the 3 town roles you offered. Sane Cop Jailkeeper 1/2 shot vigi Out of these, I'd say that I'd prefer: Cop > vigi > jailkeeper. I think all vigilantes should probably be 1 or 2 shot variants, but I think that role's balance is very weird and sometimes quite hard to quantify. I think I even lean toward vigilantes being solely 1 shot variants in minis, for the simple reason that upping the power of a vigilante doesn't necessarily make the role "better". To elucidate: It instead swings the role more toward the extremes. A good player already with the role of vigilante will be much better with a second shot, whereas a shitty player with the role of vigilante will be much worse. The swing is too volatile and I think that's a big problem with the role. A 1 shot vigilante, on the other hand, is still more volatile than other roles but is probably at a level that can be relatively easily balanced, since you can measure fairly precisely the impact it will have upon the game (a correct vigi shot has the effect of removing a scum and confirming a town, and an incorrect shot has the effect of removing a townie and still confirming a town-confirming two, actually-sometimes the loss of a townie can almost completely be offset by the confirmations, particularly in low count situations like 6v3, 5v2, 4v1 etc). On the other hand, if you go from parity cop to regular cop, or jailkeeper to doctor, or limited shot to full for any other role, the player's skill does not factor in with respect to the increase in power. All players stand to benefit from the additional functionality of the role. What I'm trying to get at here is that from a balance perspective you want to limit volatility. It's probably one of the biggest concerns, in my opinion, probably second only to preventing breaking strategies like follow-the-cop (which I suppose you could peg as volatile, in a sense). To this end you need to balance around vigilantes or remove them completely. In the setups I run, I always look for "worst cases" and "what does it take to win?" by both sides. In 12/13p minis, in the absence of vigilantes, this should almost always be a worst-case of 3/4 cycles to lose for both sides. With vigilantes it's probably acceptable if it's 2 cycle worst-case for mafia to lose. Thus, something like two vigilantes should be avoided. Including an SK is a different story, as with a 3P the pressure is higher on mafia and SK to figure each other out, and so there is a mutual benefit to lynching the opposite party in this scenario. It also changes the claim dynamic for vigis and SKs, as I'll get into later. The problem with there being a small number of roles is that individual players can make breaking strategies based upon the knowledge they glean from the setup. I experimented with no-knowledge RBs and other things to minimize player knowledge of the setup in my games, and I think I like it better. Why do I think this? Imagine in your case, Aquanim, that we decide to pick two roles for town. There are 3 combinations possible. Suppose now that I roll 1-shot vigi in your game. I would argue that in almost every instance, that my best play should be to claim immediately after shooting. Why, and why wouldn't I ordinarily claim immediately after shooting in, let's say, a C9++ variant that I run? The answer is simple. I as the vigi know that there has to exist another blue. 50% of the time, that blue is a JK, and now I've given him an optimal strategy. Jail me every night, and I am a confirmed, unkillable townie. If I shot a scum, this means we will almost certainly win if the mafia cannot find the JK within two cycles, and we are still very likely to win if the JK lives for one cycle, even ignoring almost everything else. The other 50% of the time, I will die, but that's okay, since I've told the cop he should not be investigating me. What if roleblocks are known? Then, I will almost certainly know what the other blue is, and I can choose not to claim if it's not a JK and I think it'll be better for me to claim when there are less townies. I would say that this is borderline gamebreaking. In the variants of normal minis that I run, the vigi after shooting would ordinarily only claim if he is going to get lynched or if he thinks that claiming will consolidate the lynch for the day onto scum, or if he thinks he is a likely target for mafia. He would not ordinarily have such a high chance of being protected and he has no way to predict what the setup will look like-thus, in most situations he'll only be confirmed for one cycle. In the above situation, there is always at least a 50% chance that there is a jailkeeper, and if roleblock information is given by the host then the setup will essentially be public knowledge. Adding roles and changing the chance that they appear, or forcing certain role combinations, will generally prevent this. The former obviously makes it harder to balance. e: also something I failed to make clear that I just noticed: my point with respect to the "what role would you prefer?" bit was to express the idea that each different combination of possible roles should ideally be as balanced as possible. If town get cop + vigi and this is universally considered better than, I dunno, jailkeeper + vigi or something like that, but mafia get the same roles regardless, then we have an imbalance that we have introduced. | ||
|
Aquanim
Australia2849 Posts
| ||
|
Aquanim
Australia2849 Posts
| ||
|
wherebugsgo
Japan10647 Posts
On January 04 2014 17:43 Aquanim wrote: I... may not have made something clear. Whether there's one or two power roles would be randomly determined each game. Whoops. Even in this case, I think you just halve the probabilities. It's still optimal to claim, I think. 50% of the time, there will be no other blue. 50% of the time, there will be one other blue. Of the 50% of the time where there is a blue, 50% of the time there will be a JK, which puts the chance for there to be a JK at 25%. Someone correct me if I did that wrong, I'm not the greatest with probabilities, sadly. | ||
|
Aquanim
Australia2849 Posts
The only way to fix that problem, if it isn't already fixed, is to eliminate vigilante as a role entirely - which I feel isn't a great solution either. | ||
|
Aquanim
Australia2849 Posts
my point with respect to the "what role would you prefer?" bit was to express the idea that each different combination of possible roles should ideally be as balanced as possible. If town get cop + vigi and this is universally considered better than, I dunno, jailkeeper + vigi or something like that, but mafia get the same roles regardless, then we have an imbalance that we have introduced. I do agree with you here, and those three roles were the closest to being both a) similar in power to each other and b) having no exceptional synergy in any particular combination that I've come up with so far. I'd love to hear better suggestions. EDIT: I think I rate jailkeeper higher than you do, and if a 50% chance of miller is added (which is probably necessary really) that nerfs the cop some. If further nerfs to cop are necessary, it could be changed into a parity cop or something. The point about swinginess re. 1 and 2-shot vigilantes seems quite reasonable. | ||
|
wherebugsgo
Japan10647 Posts
On January 04 2014 17:50 Aquanim wrote: I think your probabilities are right. That being said, I don't think there's any way around a vigilante being able to claim once they've made their shot - and I'd argue that across all the setups run on the forum, guessing that there'll be some kind of protective role is often going to be a safe bet. Far better than 25%, certainly. The only way to fix that problem, if it isn't already fixed, is to eliminate vigilante as a role entirely - which I feel isn't a great solution either. In the C9++ variant I run, it's probably about half of that. It is true that you should probably claim as a vigi immediately after shooting in most cases, but if there is a high likelihood of that causing there to be a confirmed, unkillable town, there's a problem with the setup. It's also true that in the C9++ variant I run, if you claim vigi you're not confirmed, because of the possibility of SK. In fact, SKs claiming vigi is pretty common. So no, you don't have to eliminate the vigi role. | ||
|
DarthPunk
Australia10857 Posts
On January 04 2014 17:37 Aquanim wrote: What exactly are the aspects of the GSL setup which appeal to you, out of curiosity? I don't disagree with you by any means, by the way. The most obvious unique feature IMO is that scum get very little information from the roles they get; if they get a Roleblocker town could have two power roles, or none at all. I'm not entirely sure if that's practical to replicate in a 13p game. Even in the 9p game I think there's a fair bit of difference in balance between a pcop+doctor and 2 namedVT setups (though I wouldn't venture to guess which is more balanced) I liked named VT alot as a power role because it forces you to learn how to claim properly. I like that you can sort of derive the setup after flips and claims etc. I like that you just role from several options. It is a really elegant setup and is my favourite mini setup of all time. Followed by c9++. I also like that it was created as far as I know by someone on this forum and is unique to TL. | ||
|
Aquanim
Australia2849 Posts
On January 04 2014 17:57 wherebugsgo wrote: In the C9++ variant I run, it's probably less than half of that, if not lower. It is true that you should probably claim as a vigi immediately after shooting in most cases, but if there is a high likelihood of that causing there to be a confirmed, unkillable town, there's a problem with the setup. It's also true that in the C9++ variant I run, if you claim vigi you're not confirmed, because of the possibility of SK. In fact, SKs claiming vigi is pretty common. So no, you don't have to eliminate the vigi role. The confirmed unkillable townie is a problem with any protective role in combination with pretty much anything claimable. Sure, a cop-claim could be a fake scum claim, but I've not seen that done successfully very often. (As an aside, I think the combination of vigi+JK in that hypothetical setup is 17% or so, which isn't all that high.) I'm trying to avoid including an SK so that solution to the vigilante isn't available, unfortunately. Perhaps the fear of an SK is necessary to balance a vigilante, though. | ||
|
Aquanim
Australia2849 Posts
On January 04 2014 17:46 wherebugsgo wrote: Even in this case, I think you just halve the probabilities. It's still optimal to claim, I think. 50% of the time, there will be no other blue. 50% of the time, there will be one other blue. Of the 50% of the time where there is a blue, 50% of the time there will be a JK, which puts the chance for there to be a JK at 25%. Someone correct me if I did that wrong, I'm not the greatest with probabilities, sadly. Oh bugger, I think we both failed at probability 101. tl;dr: If I'm a vigilante, that makes it more likely there's two PRs as opposed to one. There are six equally likely setups: Cop JK Vig Cop+JK Cop+Vig JK+Vig If I'm a vigilante, there are three equally likely setups: Cop JK Vig Cop+JK Cop+Vig JK+Vig So if I'm a Vigilante the probability of there being a JK as well is 33%, not 25%. | ||
|
wherebugsgo
Japan10647 Posts
On January 04 2014 18:10 Aquanim wrote: Oh bugger, I think we both failed at probability 101. tl;dr: If I'm a vigilante, that makes it more likely there's two PRs as opposed to one. There are six equally likely setups: Cop JK Vig Cop+JK Cop+Vig JK+Vig If I'm a vigilante, there are three equally likely setups: Cop JK Vig Cop+JK Cop+Vig JK+Vig So if I'm a Vigilante the probability of there being a JK as well is 33%, not 25%. Yeah lol, you're right. I knew I was forgetting something :p | ||
|
Aquanim
Australia2849 Posts
EDIT: In that case, the jailkeeper in the setup could probably be replaced with a doctor, if desired. The setup I proposed on the previous page probably isn't ideally what we want here (for starters, Mafia Rolecop is not really TL standard) but at the very least I think it's brought into focus some constraints which the ideal setup we're looking for has to satsify - particularly, that there has to be some solution to follow-the-cop and follow-the-confirmed-townie. I think if anyone else has any ideas for a standard normal mini setup, whether you think it's a particularly good idea or not, I'd like to see them. At the least, we'll work out more things which need to be avoided, mitigated or desired. At best, one of the setups could be chosen or modified by consensus as the standard. | ||
|
Blazinghand
United States25558 Posts
some notes on 10-3 A worst case for town would be like D1 10-3 D2 8-3 D3 6-3 D4 4-3 LYLO Giving town 3 mislynches before a loss. Contrasting the 7-2 setup, this is one extra mislynch. On the other hand, scum has more freedom with their shots (ie not only having to shoot blues) since there's a lower number of blue roles in the setup. Haven't given a HUGE amount of thought to this but this is probably how I'd do it. I'd consider scaling it up to 3 blue roles, or having the option for a 3rd blue role half the time or something, and maybe giving scum a certain chance to have an additional role. Ideally though the scumteam can only look at themselves to narrow down what the town's blue roles are a little bit. So an example of what i'm talking about might be like this: GSL Mini Mafia 13p 10 town, 3 scum. Roll 1d8 and select setup from below. Doctors are non-conseq, RBs inform the target even if it has no power, Pcops can't target themselves. A: 8x VT, 1x Pcop, 1x Doc, 1x Goon, 1x RB, 1x GF B: 8x VT, 1x Pcop, 1x BoxeR, 2x Goon, 1x RB C: 8x VT, 1x Doc, 1x BoxeR, 2x Goon, 1x RB D: 8x VT, 2x BoxeR, 1x Goon, 1x RB, 1x GF E: 7x VT, 1x Pcop, 1x Doc, 1x BoxeR, 1x Goon, 1x RB, 1x GF F: 7x VT, 1x Pcop, 2x BoxeR, 2x Goon, 1x RB G: 7x VT, 1x Doc, 2x BoxeR, 2x Goon, 1x RB H: 7x VT, 3x BoxeR, 1x Goon, 1x RB, 1x GF So what does this mean? Well, it means if you're a BoxeR as usual all you know is it's not Setup A. Unlike our normal GSL games though, there's still the possibility you have both a Pcop and a Doc on your side. You're basically in the dark and have to rely on your claim. A Doc knows he either has 2 boxers, a boxer and a cop, a boxer, or a cop as teammates, but not which. A Cop knows he either has 2 boxers, a boxer and a doc, a boxer, or a doc as teammates. What this does is give us additional scum some risky fakeclaim options. 50% of the time, there is a third blue role, and 50% of the time there isn't. Town doesn't know which setup it is, and even if scum fakeclaim blue at LYLO and get counterclaimed, there's a chance to win since it's not clear WHICH blue is lying (as was the case in GSL). Another important piece of info here is that scum know how many power-having blue roles they're up against. If they have goon goon RB, they know they're up against a single empowered blue and 1-2 boxers. If they have goon GF RB, they know they're up against all boxers or cop doc + 0-1 boxers. Info, but not PERFECT info-- enough for scum to play strategically. It's a slight buff to scum giving them always a roleblocker, and pretty importantly, if you're a cop there's a 50% chance you're up against a godfather. In this setup, there are more options. This means that scum has a slight advantage over town, since that limited information prevents things like massclaiming being quite as strong. This is probably a good thing since the 9-player GSL setup was a bit town-favored. In the case of setup H, 3 of 13 players being confirmed town on D1 is not nearly as bad as 2 of 9 players in terms of thread presence. I'll probably run this setup for my next Normal Mini and see how it works out. | ||
|
wherebugsgo
Japan10647 Posts
On January 04 2014 18:24 Aquanim wrote: Does changing the doctor/jailkeeper/protective role such that it can't protect the same person twice in a row alleviate your concerns at all WBG? I think that would mean that scum should at least get the opportunity to kill the confirmed townie at some point. It does something to solve follow-the-cop (a very related problem) as well. EDIT: In that case, the jailkeeper in the setup could probably be replaced with a doctor, if desired. The setup I proposed on the previous page probably isn't ideally what we want here (for starters, Mafia Rolecop is not really TL standard) but at the very least I think it's brought into focus some constraints which the ideal setup we're looking for has to satsify - particularly, that there has to be some solution to follow-the-cop and follow-the-confirmed-townie. I think if anyone else has any ideas for a standard normal mini setup, whether you think it's a particularly good idea or not, I'd like to see them. At the least, we'll work out more things which need to be avoided, mitigated or desired. At best, one of the setups could be chosen or modified by consensus as the standard. Yeah, it would, but I'm personally not a fan of including restrictions on how people can use their roles in balancing. For the same reason I don't like telling vigilantes they can't shoot on n1, though I think I experimented with things like that in the past. e: wtf is a boxer | ||
|
DarthPunk
Australia10857 Posts
On January 04 2014 18:39 Blazinghand wrote: GSL setup for 13 players (10 town 3 mafia) is tougher. some notes on 10-3 A worst case for town would be like D1 10-3 D2 8-3 D3 6-3 D4 4-3 LYLO Giving town 3 mislynches before a loss. Contrasting the 7-2 setup, this is one extra mislynch. On the other hand, scum has more freedom with their shots (ie not only having to shoot blues) since there's a lower number of blue roles in the setup. Haven't given a HUGE amount of thought to this but this is probably how I'd do it. I'd consider scaling it up to 3 blue roles, or having the option for a 3rd blue role half the time or something, and maybe giving scum a certain chance to have an additional role. Ideally though the scumteam can only look at themselves to narrow down what the town's blue roles are a little bit. So an example of what i'm talking about might be like this: GSL Mini Mafia 13p 10 town, 3 scum. Roll 1d8 and select setup from below. Doctors are non-conseq, RBs inform the target even if it has no power, Pcops can't target themselves. A: 8x VT, 1x Pcop, 1x Doc, 1x Goon, 1x RB, 1x GF B: 8x VT, 1x Pcop, 1x BoxeR, 2x Goon, 1x RB C: 8x VT, 1x Doc, 1x BoxeR, 2x Goon, 1x RB D: 8x VT, 2x BoxeR, 1x Goon, 1x RB, 1x GF E: 7x VT, 1x Pcop, 1x Doc, 1x BoxeR, 1x Goon, 1x RB, 1x GF F: 7x VT, 1x Pcop, 2x BoxeR, 2x Goon, 1x RB G: 7x VT, 1x Doc, 2x BoxeR, 2x Goon, 1x RB H: 7x VT, 3x BoxeR, 1x Goon, 1x RB, 1x GF So what does this mean? Well, it means if you're a BoxeR as usual all you know is it's not Setup A. Unlike our normal GSL games though, there's still the possibility you have both a Pcop and a Doc on your side. You're basically in the dark and have to rely on your claim. A Doc knows he either has 2 boxers, a boxer and a cop, a boxer, or a cop as teammates, but not which. A Cop knows he either has 2 boxers, a boxer and a doc, a boxer, or a doc as teammates. What this does is give us additional scum some risky fakeclaim options. 50% of the time, there is a third blue role, and 50% of the time there isn't. Town doesn't know which setup it is, and even if scum fakeclaim blue at LYLO and get counterclaimed, there's a chance to win since it's not clear WHICH blue is lying (as was the case in GSL). Another important piece of info here is that scum know how many power-having blue roles they're up against. If they have goon goon RB, they know they're up against a single empowered blue and 1-2 boxers. If they have goon GF RB, they know they're up against all boxers or cop doc + 0-1 boxers. Info, but not PERFECT info-- enough for scum to play strategically. It's a slight buff to scum giving them always a roleblocker, and pretty importantly, if you're a cop there's a 50% chance you're up against a godfather. In this setup, there are more options. This means that scum has a slight advantage over town, since that limited information prevents things like massclaiming being quite as strong. This is probably a good thing since the 9-player GSL setup was a bit town-favored. In the case of setup H, 3 of 13 players being confirmed town on D1 is not nearly as bad as 2 of 9 players in terms of thread presence. I'll probably run this setup for my next Normal Mini and see how it works out.  Blazinghand | ||
|
Blazinghand
United States25558 Posts
On January 04 2014 18:39 wherebugsgo wrote: Yeah, it would, but I'm personally not a fan of including restrictions on how people can use their roles in balancing. For the same reason I don't like telling vigilantes they can't shoot on n1, though I think I experimented with things like that in the past. e: wtf is a boxer BoxeR is a blue with no powers. His only power is he receives a PM that looks different than the PMs the VTs get. Because the setup is semi-open, he can claim and has some setup info. btw anyone who wants can use that setup of suggest changes. not sure how it would work yet. | ||
|
DarthPunk
Australia10857 Posts
| ||
|
Blazinghand
United States25558 Posts
also since they're not aware of each other you can have 3 boxers in a setup without breaking it lol | ||
| ||
