ESPIONAGE: A brand new simple forum game! (Central Rules & Information Topic)

[Jd-]

New Ability Added: "Assassinate" for Informants

A new ability, Assassinate, has been added to the game as a standard feature. This ability allows Informants to issue a one-time-use kill, which is specified in the results for a given phase as the target having "died from an accident" (thus signaling to the player population the ability is now exhausted). However, this ability may only be used ONCE total, no matter how many Informants are present in a game. If multiple Informants use the ability in a given phase, whichever sends the action last is considered the valid use with the rest being canceled. Should the Spy and Informants target the same player for a given phase with their respective kill abilities, the Spy's ability takes precedent and the Informants' Assassinate ability is canceled. The Spy may use the Order system to help coordinate kills and suggest who they would like the Informants to take out.

[Jd-]

Additional point to raise: Should we change or remove the rule about revealing partial lists in death? Should we leave it the same? Depending on if we go with the last proposal or not, we may need some changes to this. If we do make it so it's fair game to post the new versions of lists, we maybe should scratch dead list reveals? Or, leave it in? I'm not entirely sure, but thoughts are of course welcome.

[bash7353]

If the lists are generated the way that it was said they were during Round 1 then every time the GM sets up someone's list, there's a chance of 14/57 that only one of the special roles in in the list, 7/57 that both are and 22/57 that none of them are. If the latter happens the list is regenerated until the results one that doesn't include only Civilians, so that gives us those figures:

Spy, but no Detective on the list: 40%
Detective, but no Spy on the list: also 40%
Both on the list: 20%

Those numbers would be different if lists contained more or less names.

For the expected value for how often a name would appear in lists you'd get: 10.8 for both Detective and Spy and 5.8 for every Civilian. That essentially means if you let a GM generate lists for everyone in this game with Raiden as Spy and Kleene as Detective a lot of times, how often would a name appear on average.

Note that 10.8 is more twice as big as 5.8, Detective and Spy appear that much more often in the lists, that's something a Civilian that gets a hold of entire lists from other players could realise and help narrow down things. But from a statistical point of view Spy and Detective will always have the same likelihood. So maybe with enough lists they can say, Oh, Player X is probably a special role. But they wouldn't have anything to judge whether Detective or Spy is more likely.

If a Spy gets a random list that he finds himself on, that list is absolutely useless. There's no conclusion that could be drawn from this at all. The people that appear with him on the list are not more or less likely to be the Detective. However, if he receives a list that doesn't contain himself, he obviously knows the Detectives is among them. A second list like that and it gets very narrow. But, if you give the Spy a random list, there's a 40% chance for him not to appear himself, so on average the Spy would need 5 random lists, so that two wouldn't include himself.

The above stated is also true for the Detective receiving lists in trying to find out who the Spy is.


All of what I'm saying here changes when we introduce Informants who don't get any sort of list, or change how many names there are on a list at all. Also, when the number of total players changes everything's different.

[bash7353]

New List Proposal: Weighted Totals

One idea for solving the list problem: It is no longer 100% guaranteed your list contains the Spy or Detective. Each list would have, say, 7 names on it, but it would no longer be guaranteed that the Spy/Detective are on it. What may well solve this entire thing is to weight the roles. Let's say that there's only a 35% chance that your list contains the Spy, but there's a 65% chance it contains the Detective. This prevents people from just posting their lists outright at the start of the game, because the Detective will definitely be in serious danger if all those lists were made public. And yet, even at the end, if everyone did post their list, it's no longer guaranteed the Spy would appear on the lists that may or may not have contained the Detective. Since their inclusions will be independent of one another, it's impossible to know if one, both, or either would be on there from the start.

65% and 35% added up equal 100%, is that on purpose? 'Cause you seem to imply that it's still possible that both Detective and Spy are on the list, and impossible for neither.

[Jd-]

New List Proposal: Weighted Totals

One idea for solving the list problem: It is no longer 100% guaranteed your list contains the Spy or Detective. Each list would have, say, 7 names on it, but it would no longer be guaranteed that the Spy/Detective are on it. What may well solve this entire thing is to weight the roles. Let's say that there's only a 35% chance that your list contains the Spy, but there's a 65% chance it contains the Detective. This prevents people from just posting their lists outright at the start of the game, because the Detective will definitely be in serious danger if all those lists were made public. And yet, even at the end, if everyone did post their list, it's no longer guaranteed the Spy would appear on the lists that may or may not have contained the Detective. Since their inclusions will be independent of one another, it's impossible to know if one, both, or either would be on there from the start.

65% and 35% added up equal 100%, is that on purpose? 'Cause you seem to imply that it's still possible that both Detective and Spy are on the list, and impossible for neither.

Nah, not really. It was just a way of attempting to introduce a mechanic that gave reason to not post the lists publicly immediately (the Detective would be more easily identified, theoretically, and die at the hands of the Spy as a result). Meanwhile, the Spy would not be as easily identified, theoretically.

As I said in the post, the likelihood of one or the other being in the list would be completely independent. There'd be a 35% chance of the Spy assuming one of the 7 slots. All the same, there would be a 65% chance of the Detective assuming one of the 7 slots. Meaning both could show up, neither could go up, or just one or the other could appear.

My worry was the reverse engineering involved. Would it be possible to locate the Spy in an end-game situation? After the Detective is dead, would it be possible for a reliable calculation to be made locating the Spy based on everyone posting their lists?

[bash7353]

Okay, I understand what you mean now. I'll try to modify the math about what I posted about how things went statistically in the actual Round 1 to what would have happened if we had played with what you're suggesting.

[Kleene Onigiri]

Do I understand it correctly or not?
Does the spy and or the detective have to be on one on the list? Or is it possible that none of them are on the list at all?

Because then this part from googleearth:
"However, if he receives a list that doesn't contain himself, he obviously knows the Detectives is among them. A second list like that and it gets very narrow. But, if you give the Spy a random list, there's a 40% chance for him not to appear himself, so on average the Spy would need 5 random lists, so that two wouldn't include himself."
wouldn't actually be true. Because it's possible that there is no spy or detective on that list at all.

[Jd-]

For additional clarification: Neither the Spy nor the Detective have to be on the list in this idea.

Ultimately, if we settle on this idea, I'll talk with some people about creating a generator for us that can be customized to immediately create all the lists based on whatever inputs we'd like. It should make GMing a lot easier.

[Fujiwara]

Do I understand it correctly or not?
Does the spy and or the detective have to be on one on the list? Or is it possible that none of them are on the list at all?

Because then this part from googleearth:
"However, if he receives a list that doesn't contain himself, he obviously knows the Detectives is among them. A second list like that and it gets very narrow. But, if you give the Spy a random list, there's a 40% chance for him not to appear himself, so on average the Spy would need 5 random lists, so that two wouldn't include himself."
wouldn't actually be true. Because it's possible that there is no spy or detective on that list at all.

In googleearth's first post, he was talking about the probabilities in the actual round 1, and not about Jd's proposed modification. I was a bit confused by that at first as well. In the actual round one it was guaranteed that at least either the spy or the detective were on the list, in the new proposal this is not the case.

[bash7353]

Okay, here's what would change:

Neither Spy nor Detective on the list: 22.75%
Spy, but no Detective on the list: 12.25%
Detective, but no Spy on the list: also 42.25%
Both on the list: 22.75%

That gives us interesting expected values:
Detective: 11.7
Spy: 6.3
Every Civilian: 6

If every list would be posted we couldn't tell who the Spy is by merely looking at the number of times the Spy appears. The expected values for a Civilian and the Spy would be too close for that. Identifying the Detective would be pretty easy though. There'd be one name that appears almost twice as often as all the others.

The big difference between the suggestion and how we actually played is there's no certainty anymore. Before, we could look at the list and if the Detective isn't on it, well then we know so much more than people who do have the Detective on the list. That's why lists become less powerful after the Detective's identity has been revealed if we play like that compared to how powerful some lists became in Round 3.

That means, though, that the lists will only be useful when trying to identify the Detective, which is somewhat important to know even for a Civilian, but I fear that people would just disregard the lists altogether 'cause they're no help at all when it comes to deciding who to vote for. And what's the point of lists people won't use?

Looking at what I've posted earlier about the actual Round 1, I think there already is incentive to keep your list secret throughout the game. As soon as the Detective dies that all changes, though. But then we will have the Spy and potential Informants who are still in the game who will pretend to have lists as well.

[Jd-]

What if we raise it to 40%? 45%? Would that make things make a little more sense?

[Jd-]

OK, I know this is going to be a bit controversial after the conensus we reached yesterday, but maybe it's the answer: Informants.

What if Informants are included in the lists now as a possibility? Keep in mind that in the last round, there was really just "one" Informant in that we were one team. In the real game, there will be up to 3. So, what if we included them as one of the possibilities and we returned it to having a guaranteed special role in your list?

This opens up all kinds of questions. Foremost, the Informants are now in danger, so they have to be careful with how they defend their Spies (even more than usual). In fact, a really keen civilian may catch on and monitor them for a few phases without exposing them, just to see if they can get a read on who the Spy is.

With this system, we'd adjust the numbers a bit, but return it to "someone with a special role is on your list". However, I still want to weight it in the Detective's favor, purely because of the end-game.

Is this the answer to this issue? We now have two distinct possibilities on the table, both with pros and cons. I'd like to hear opinions from all the regulars we can get in here.

[Jd-]

So, if we use the above idea (introducing the Informants to the lists), let's talk how this works. Numbers wise, let's say... 55% Detective, 45% Informant, 35% Spy. The big difference being that one role must be included on the list. If I'm thinking about this correctly, it would be very, very, very disadvantageous to post all the lists anytime before the end-game. At that point, there would be a lot of confusion and hopefully no reliable means of determining who the Spy is once access to all 15 lists are in the public.

Let's again use the Round 1 list for ease, and then just randomly add in three Informants to maximize things:

---

Jd-
PhoenixTears (Informant)
dumytru
Yuri Iwamoto
Kleene Onigiri (Detective)
Commi-Ninja
Meme
Stopwatch (Informant)
Kamite
Monsi
Raiden (Spy)
shinichi'sapprentice
googleearth (Informant)
Nix
Conan-chandesune
miyano_shiho
Raifuujin
kkslider5552000
KainTheVampire
aly_angelflight

[Jd-]

I've got someone on board to create a web app for easy generation of the lists, so that the GM has to do very little and every list will be calculated accurately. Of course, before he starts on it, we'll have to nail down what we want the program to do, so let's talk it out here.

PLEASE NOTE: From here on, lists will be publicly shareable at the civilian's discretion.

---

Possibility #1:
— Lists are now determined by weighted totals BUT do not have to contain any special role players (Detective NOR Spy). There would be a weighted total for each appearing, such as 65% chance of a Detective appearing, or 35% of a Spy appearing. However, neither may appear as well, leaving only a list of civilians and possibly Informants. These weights are independent, meaning the Detective and Spy could both appear.
— The end-game is now protected, because once the Detective dies, people posting all of their lists at once won't identify the Spy immediately unlike how it'd normally be.
— Some civilian lists will contain no special role players. This means some civilian lists are of lesser if little strategic use.
— Informants are considered civilians for calculation purposes, in that their appearances in lists is as a civilian and their informant status is not considered.

---

Possibility #2:
— Lists are similar to how they were before but now include Informants as one of the roles included in lists.
— The lists are determined by weighted totals (ex. 55% likely to have a Detective on your list, 45% likely to have an Informant, 35% likely to have a Spy) BUT must contain at least one Detective, one Spy, or one Informant. These weights are independent, meaning the Detective, Spy, and Informants could all appear or any combination therein.
— The end-game should be somewhat protected (as in, once the detective dies and everyone begins posting lists freely), but maybe not entirely. We need some numbers crunched on this to be sure.
— Most games will include 1-2 Informants, so keep that in mind.

[Fujiwara]

Possibility #1:
— Lists are now determined by weighted totals BUT do not have to contain any special role players (Detective NOR Spy). There would be a weighted total for each appearing, such as 65% chance of a Detective appearing, or 35% of a Spy appearing. However, neither may appear as well, leaving only a list of civilians and possibly Informants.
— The end-game is now protected, because once the Detective dies, people posting all of their lists at once won't identify the Spy immediately unlike how it'd normally be.
— Some civilian lists will contain no special role players. This means some civilian lists are of lesser if little strategic use.
— Informants are considered civilians for calculation purposes, in that their appearances in lists is as a civilian and their informant status is not considered.

Personally, I think this version could work out well. In the previous rounds, people were tempted to play bingo with their lists, as someone formulated it, and I don't think that's what the game should be all about. Without the lists to help us, we'll have to actually deduce the identity of the spy based on what people are saying as well as the spy's actions. The most important function of the lists will be that they can help us protect the detective: There's a higher chance of voting out the detective if we vote from the lists, so we should preferably vote for someone not on our list instead. Conversely, we'll get no hints about the spy's identity through voting, but the spy might be able to narrow down the potential detectives through the voting results.
One thing we should consider with this variation is that it gives the spy an advantage over before. If we also include the new ability 'assassinate,' the spy side might get so strong if there are too many informants - the 3 informants you mentioned sound like a quite a lot. Let's not forget that due to Conan-chandesune's early demise in round 2, we've never actually played a round with more than 1 informant, so if we suddenly get no real information through the votes, a new informant ability and 2 or more informants, it might not be very balanced either ... but, well, I guess testing these things out in practice should give us a good idea.

Possibility #2:
— Lists are similar to how they were before but now include Informants as one of the roles included in lists.
— The lists are determined by weighted totals (ex. 55% likely to have a Detective on your list, 45% likely to have an Informant, 35% likely to have a Spy) BUT must contain at least one Detective, one Spy, or one Informant.
— Most games will include 1-2 Informants, so keep that in mind.

First off, is it 45% for each informant (if there's more than 1), or 45% for any informant?
The problem with this option is that it's very similar to the orginal version, so the point of contention is: What happens when the detective is out of the game? At this point everyone can post their lists and for every list that doesn't contain the detective, we’ll know that it contains either the spy or the informant (or both).
So the probablilities P(spy) and P(informant) are actually quite a bit higher than 45% and 35% respectively on these lists, since the combined probability P(spy or informant) has to be 100%. In fact, even the base probabilities of 55/45/35 are not the actual probabilites, since the redraws to ensure at least one special role is on the list increases each of these probabilities. I’m not sure how to calculate the actual probabilities in this case – maybe someone else knows? If not I'll research the matter and get back to you later.