Design Review: 9.0 Design Changelog
04-21-2013, 07:30 AM (This post was last modified: 04-21-2013 10:17 AM by Chocolate Pi.)
Post: #1
Design Review: 9.0 Design Changelog
Warning: This post is absurdly long! It covers the design process for every change made from WH 8.2 to 9.0.

14 months ago, Gambler and Recluse had a miller side-effect, and there existed Vigilante and Jester roles with non-trivial state. Last year, when I left for GDC, I considered these good roles. Not the best, and in fact probably the worst--but something had to be the worst, right? Worst could still be good. But when I got back from GDC, they were no longer good.

The bar had been raised when I wasn't looking.

I keep a spreadsheet of every published change I've made to WH over the last 6 years. There's a trend, a growing sense of the minimum standard of quality that content needs. When I left for GDC this year, if you asked me what the "worst" designed role was, I'd probably say Angel Worshiper or Oracle--shortly before clarifying that I still thought they were good content for the game. But when I got back from GDC, they were no longer good.

The bar had been raised when I wasn't looking.

But this was a little different, more urgent. I wanted to publish soon. That meant this was the last chance. I needed to solve not just these problems, but also next year's, and every year's. The game doesn't have to be perfect, but I wanted it to be darn close. I needed to raise the bar, and then raise again until I couldn't any more.

I've been diagnosed with having a "depth-first-search problem." In other words, left to my own devices I operate by focusing very intensely on a single problem. This is generally a bad thing, but can be useful for solving unusually "dense" problems. When confronted with a particularly nasty problem that I can willingly DFS on, I go into what I call a "design coma"--generally until I solve it. The tougher ones can last an hour or two.

I've spent most of the last three weeks in a Witchhunt design coma. I'm physically and mentally wiped out, but hell if it didn't get results. I'm extremely satisfied with the state of literally everything at this point.

The design work is decisively done, and it feels fantastic.

Angels and Demons
  • Angels were a tad too powerful than we'd like, just barely.
  • Dead Witches had basically nothing to do--the "Demons" added to online games were weak to the point of triviality, added state, and couldn't really be used offline.
  • Dead witches voting with Angels confused people to no end.
  • Angels felt bad wasting protects, and smart Angels began abstaining early votes.
  • Witch Angels would try to waste protects and accidentally block kills.
  • Keeping track of angel protects was annoying to moderate (state!) and confusing/frustrating for the Angels themselves.
  • Players disliked the restriction of Village Spies not being able to be protected by the Angels, and often forgot about it.
  • Players often forgot that Angel protection only covered Witch kills, which was unequal to the effects of roles like Gambler/Recluse.

In short, Angels had issues. These all paled in comparison the dozen great things they were doing for the game, but there was tons of room for improvement and streamlining.

We are taking away virtually all the rules limitations on Angels. Instead, we are going to let the Demons--the dead Witches--limit them.

Every night, Demons can veto one target. Angels can protect anyone else, from all kills that night.

And that's it.

All dead Villagers are Angels, all dead Witches are Demons. There is no state. Every single one of the listed problems is addressed. Angels are still powerful, but a little less; specifically, they are way more powerful until a Witch dies, and considerably less powerful after. Holy claims in particular are hurt, which is welcome. The new concerns are that this system is too degenerate and weak once Witches stick to a Demon-sync'd plan.

The solution is to give the first dead player(s) an Archangel card.

[Image: aa_Archangel.jpg]

This is a resource he can spend to override and ignore the Demon's veto one night, giving Angels a critical ace in the hole. Witches that are in sync with the Demons will be telling Angels exactly who and when to Archangel, so that's cool.

For awhile, the plan was for each expansion to have their own unique Archangel card. On top of the dubious benefit of such added complexity, there was really nothing that offered gameplay quite as compelling as the "base" Archangel. Because these cards were meant to be resources and not linked to player identity, it was acceptable to use multiple identical copies of one Archangel.

But what if a Witch dies first? For some time, the solution was just to let them take the Archangel card, obviously never use it, and just sit on it cackling. Some players wanted Demons to get their own cards instead, but nothing we give them could be as meaningful as taking away an Archangel card.

[Image: aa_Archdemon.jpg]

Enter the double-sided Archangel card; flip it over and it's an Archdemon card. This can be spent to veto an extra target. This is a nice bonus but not terribly important or powerful; however, there is one case where it is very potent, and we'll get to that.

The current plan is to give out one Archangel card for every 10 players in the game, rounded down. So, a 24-player game would give Archangel cards to the first 2 players announced dead.

The idea in all of this is three-fold:
  • We want to keep the Angel interaction with Demons (and the Demon interaction with the Coven) gameplay as dynamic as possible.
  • We want the first dead player(s) to have a consolation prize, to feel empowered and important rather than the usual awful experience of dying first in a mafia game.
  • We want to amplify the rubber-banding effect that angels already provide, minimizing the impact of the first lynch (appropriately) in order to emphasize the later actions.

It's obviously important that we not overshoot the last point, but I'm pretty sure we are quite safe. Getting an Archdemon instead of having to deal with an Archangel is pretty nice for Witches, but is it ever truly worth being down an entire team member, from the start of the game? It's a very nice consolation for losing someone to the first lynch, but not much more.

Because we want to keep dead player participation optional, Archangel cards can be freely given to other Angels/Demons. We just can't let living players see them, because they cannot be allowed to know if/when they have been used.


Angel Worshiper was lame.

[Image: Angel%20Worshiper.jpg]

It was the only role with two truly separate (if thematically similar) abilities, both of which were built into the old Angel paradigm. There was no real gameplay with the first, and it wasn't even that powerful--it was also nebulous and confusing to the Angels themselves in most cases--why do they care if they can protect some random, low-power neutral multiple times? It's a bit of a catch-22 when his power was that you could choice to save him multiple times if he was powerful. His Halftime ability was very hit or miss, and interact with Angels in a potentially destructive and controlling way. Worst case, the information was meaningless and useless. Best case, you just town-confirmed an extra 3-4 people, effectively ending the game. It was way too swingy while being generally weak. On top of all of this, scientific investigation proved that Angel Worshiper is a truly stupid name.

So what to replace it with? The bean counters (aka me) said that the Halftime expansion needed another kill (or ~half) to match the average pacing of the other two, so we wanted a kill role. And boy, Assassin was doing pretty good, although a bit weaker than originally thought, at least compared to Hunter and Entertainer. It turns out, the original reason I investigated Angels and Demons as the first new Angels paradigm was because I wanted to make the following character:

[Image: 21_Summoner.jpg]

I really like this character. It's a very cool Angel interaction, because it leverages the Demons as an antagonistic factor to allow Angel interaction without any messaging integrity.

What does that mean? Well, the main risk with Angel interaction is the Angels--somehow, anyhow--communicating with town with their nearly unlimited knowledge. It's easy to see that even a single bit of data, a simple yes-or-no answer to a question, is game-breaking.

But here, there is no way for Angels to send a clear message at all. The Demons can ALWAYS sabotage communication with their equivalent, indistinguishable action.

So generally, it's a pretty cool concept for a kill--Angels save the most pro-town guy, Demons save the most anti-town guy, it kills the guy "in the middle". Everyone I talked to liked the idea, but this bumped into the same problem as the Assassin.

It was a kill with most of the power taken out of the hands of the character making it, which makes the character weaker. For example: for town the Assassin is just a single kill (with the slight advantage of Spy oversight for non-Spies), and for Witches it is just a single kill of their second-most preferred target. There's a slim chance of Assassin being a double kill, but it's really unlikely. Either way is way worse than just your baseline Hunter.

Summoner had the same problem, but worse. Not only did Witches kill their runner-up, but now town did too! This meant town had to guess two Witches to kill one! The role was overall very weak and unsatisfying as the Summoner himself, even if we liked the framework.

What if we we changed "must save" to "may save?" The roles would suddenly get WAY stronger, but would it be an appropriate amount?

The Assassin would essentially become a double kill for town (with Spy supervision) and a single-potential-double kill for Witches, double if they tricked the Spies into signing off on it. It's clearly the best town kill role and worst Witch kill role, but because our balance criteria is average magnitude, that's fine. "May" Assassin is okay and a welcome solution.

What about Summoner? Town still has to guess 2 witches to kill 1, but can also now guess 3 witches to kill 2! (And still enjoys benefits of limited Angel oversight, so they don't blow up the Priest.) Witches get a double kill, but it's still limited to their #2 and #3 targets. That's still a little strong, but remember: Angels are making this decision with knowledge of who they are protecting that night--from all kills! If any of the 3 Witch Summoner targets were protected by the Angels, they can just choose to save another and effectively block two of the kills! So while Witches get a baseline double kill, it's not on their most preferred target and is sometimes reduced to a single kill. Summoner appears to be on average the worst town kill role and best Witch kill role. This works.

And thus, Assassin and Summoner were made into optional saves.

An Aside On Complexity

Let's talk about perceptions for a second.

There are three big categories in game design where player's perception of a game can sharply diverge from their actual experience in problematic ways:
  • Complexity
  • Understanding
  • Balance

I've already written about how perceptions of balance are rarely aligned with actual balance as measured by metrics. We are quick to intuitively label "balanced" anything with easily observable pros and cons, even if they are unequal in magnitude upon investigation. Widespread attitudes of balance or imbalance in any given competitive multiplayer game become more of a pariah built upon social phenomenon than anything else.

Perceptions of understanding can also vary from actual understanding. This isn't surprising, but has important ramifications to game design and is a good contrast for the last one:

I like to say that games have a finite complexity budget; players all have a finite amount of complexity they are willing to put up with. As a party game for massive groups of players (who may be approaching the game with wildly varying attitudes, such as getting dragged in apprehensively), Witchhunt has a shoestring complexity budget. But in many ways, the important budget is not true complexity as much as perception of complexity.

When a person decides that a game is too complex, they sort of shut off and resist absorbing additional information; they have cast a judgement and decided that the game is not something they want to invest in. The game has failed before it has begun being played. The key though, is that this is their perception of complexity. I've seen people declare some game to be "too complicated", and leave to play League of Legends, Starcraft, or Magic the Gathering! I've done it myself!

The coin has two sides. Some players will tell you those hyper-complex games are "not really that complicated" or "actually easy to learn."

It is as much a responsibility of game design to address perceptions of complexity and understanding as it is the reality of both. If a player feels a game is too complicated, if they feel intimidated and uncomfortable by it, it is a point of design failure.

Witchhunt has struggled s a lot with this. Objectively the complexity of Witchhunt is "medium-low"; we go to great lengths to keep it low, avoiding many categories of dangerous, high-cost mechanics. One a complexity scale of 1-10, where 1 is Chess and 10 is DotA, I'd peg Witchhunt around a 4; the scope and density of rules in a typical Witchhunt game is slightly below that of say, Settlers of Catan.

Where Witchhunt struggles is that most of its rules complexity is both dense and front-loaded; ~75% of the rules are all on the unique character cards, which everyone wants to look at first because they are the cool and sexy part. This complexity binge without taking the time to chew can quickly overwhelm, especially when reading all the cards. (Including ones that aren't even in the game they will be playing.)

What is the solution to this? Well, there's no silver bullet. But it's still a great introduction to changing the Village Lovers.

Village Lovers

Village Lovers was the most complicated sub-system in Witchhunt.

This was a pretty objective fact; the sub-system consisted of 166 words spanning 5 different cards. All of this was very dense yet lacking in any structure. The entire skeleton was stable and offered sound, interesting gameplay, but very nebulous conceptually. Many interactions Lovers had with other cards were not intuitive (or at least not obvious), with sometimes bizarre ramifications.

Why have Lovers at all? Because players love them, fittingly enough. They are top-tier, universal drivers of narrative. The idea of two random players being lovers is the rare type of joke that never gets old. Players all just really like Lovers.

I surveyed a large number of players in multiple contexts with a single question: What do you think the most complicated sub-system in Witchhunt is?

Exactly zero said Lovers. (The most common answer was King's Court; it probably is the true runner-up, but it is much more structured and organized conceptually.)

When I pointed out that Lovers was the most complicated part of the game, virtually everyone disagreed. I asked if they understood Lovers, and almost everyone said of course. I'd ask them to explain Lovers and their mechanics and base interactions to me off the top of their head. Not a single person was able to do this correctly. Their perceptions of understanding and complexity did not match reality.

On top of this, we had non-trivial mechanical concerns:

[Image: Cupid.jpg]

Cupid was in the awkward position of having a big chunk of pro-town power put on a neutral role, made possible by having it apply regardless of his own alignment. This was extremely interesting, and also extremely weird.

[Image: Mercutio.jpg]

Mercutio was a stable but purposeless role. Lovers knowing who Mercutio--some random neutral role--was proved to be useless and borderline confusing information. ("Okay great, how does that help us any?") Mercutio's half-supervised kill was a tad weak compared to his offensive role peers, and the Lover suicide trigger turned out to be unsatisfying and felt outside his control.

Also, as everyone was quick to point out: "Mercutio didn't actually kill anybody..."

We needed a new Lovers paradigm that was much simpler.

Lovers are intended/needed to be power-towns. Bifurcation sort of demands this; a pair of lovers is best considered balanced wise as a double-headed town-confirmed player. They are defined by having a suicide pact, which is a harsh drawback--a rough starting point for designing power-town! We have to come up with some crazy advantage, particularly a defensive advantage, that is carefully measured to push them to just the right level of power and threat.

I kicked around a ton of different ideas. It seemed I could easily get only two out of three: strong, simple, and good gameplay. Strong, simple, but degenerate Lovers were easy. Strong, interesting, but complex Lovers were also easy. Even simple and interesting Lovers with inappropriate power levels were easy. It was getting all three that was hard.

Ultimately, the new angel paradigm enabled the solution:

[Image: _Village%20Lover.jpg]

The magic here is that Village Lovers are essentially veto-proof town-confirms that can be protected by the angels indefinitely.

Why? Because Demons can normally only veto one target. Angels can always get around the veto by just protecting the other one. (This is where Archdemon cards--vetoing an extra target--are most valuable.)

Lovers become sort of a "default option" for Angels. They can protect them indefinitely, if and only if they are willing to forgo protecting anyone else ever. But since their decision is double-blind with the Coven's, they are tempted to sneak away to protect others, and the Coven is tempted to call them on it. It creates a "yomi" situation that is exactly what we want.

Note that usually adding town-confirmed players scales town power super-exponentially. However, while this is still true here, adding a second set of Lovers has nowhere near the impact of the first. After all, Angels can only protect one pair each night! A second pair of Lovers does not want to claim when the first is publicly known, and neither probably wants to be the first to claim to begin with. (In case the other needs to)

Now we can streamline our rules text, since all of our night defense is the same--covers all kills. Gambler, Recluse, Hunter, and the new Cupid (more on him later) had their "cannot die" text converted to "protected", allowing the Lovers to share protection from those abilities. This is a great example of something that feels WAY more powerful than it is. It feels like you are getting twice your money's worth, but when you view the Lovers as a single two-headed player, it's mostly the same. When you view the Lovers as a single two-headed player who is already the default angel target, it seems downright redundant and doesn't afford them much additional safety beyond the first night. Gambler/Recluse Lovers seems insane, until you realize that this is no different from the Angels protecting them every night anyway.

So how are Witches supposed to kill them?
  • Like Bishop, the ultimate preferred strategy is overrun. If the Witches just don't die, unkillable roadblocks don't matter. Enough Witches never getting lynched should always be an unbeatable strategy.
  • It's hard, but Witches always have the potential of leading a lynch against the Lovers. It's hard since they are pseudo-townconfirmed and Witches have Cupid to deal with, but it's always *possible*.
  • An Archdemon card can let Witches take a shot. It is foiled by character protections and Archangel cards, but overall it's still a potent, universal option.
  • The most reliable tactic though, is killing them with *any* counter role. Judge, D.O.B., Loose Cannon, new Moonlight Pagan, new Twilight Pagan, and any Musketeer can all blow up Lovers. Templar also works if either is a revealed Pagan, and Wizard can also kill him. The first three are in the base set, and all the others except Wizard are in the same expansion as the Lovers.

The odds of Witches having at least one of the listed counter characters is insanely high (as high as 98% in some game sizes), so Lovers are incentivized not to claim arbitrarily. (Especially the first day, when there is no guarantee there will be an Angel that night! This plugs the "King's Court stacking" D1 claim strategy.) But don't they want the Angels to know who they are? There's a role for that.


[Image: 37_Cupid.jpg]

Knowing who the Lovers are (including both pairs, if there are two) is pretty good gameplay that we like. But we are really leveraging this as a sneaky way to show the Lovers to the first dead player!

This ups the ante on the first dead player even more, but it's still within bounds. The only time Witches would seriously consider an intentional sacrifice play D1 is in those 2% of games where they have no Lover counter. (They still have to figure out who the Lovers are themselves though; Witch Cupid is valuable for that.)

Cupid who just learns info is still a little weak on his own as a character, and is especially mediocre as a Lover himself. (The security of knowing Witches can't have Cupid is night as a Lover, but nothing amazing and doesn't feel satisfying.) How can we fix that? "What's something that everyone would like, but Lovers most of all?" Protection night 1; it's the one "gap" in the Lovers' armor. And everyone likes night 1 protection! (Witches don't care much, but Cupid is already great and cool for them.)

But we still need to replace Mercutio with a more wholesome multi-Lover-interaction character that is simple, and adds gameplay under the new system. We know


So Lover protection splashes. What if it kept splashing?

[Image: 523813_284272175015591_1173942482_n.jpg]

[Image: 43_Harlot.jpg]

This is a really cool info/defensive role, and one of the only ways psuedo-Angel-info can work. It also has a lot of narrative potential--it's easily remembered.

You can think of it as a weaker Cupid, who only learns Lovers gradually and conditionally. But she also enjoys protection, and gets clues as to the Angels' protection behavior as well as a limited idea if they are Gambler/Recluse/Cupid. It's a very interesting position.

Lover interaction roles have a lot of trouble working as Lovers themselves. Lover Harlot is not only functional, but powerful, hilarious, and awesome. It allows the protection to splash to all 4 Lovers! Like other protection roles on Lovers, this seems and feels more powerful than it is, which we love.

There is a little concern with adding another "half-night action", but Harlot should process faster than any other night action/info role. It's a worthy investment.

On the subject of Angel communication, there is potential for Harlot's information to be exploited to receive messages from Angels. Can you spot the problems with this plan though?
  • It requires publicly claimed Lovers.
  • The Harlot could be already dead.
  • The Harlot could be a Witch.
  • The Harlot could be killed, in select cases.
  • Demons can block any single target, so all tests must be done in pairs--an entire set of Lovers.
  • An unopposed Archdemon can still sabotage the message even then.
  • And above all else, constructing a message protocol out of deliberately exposing otherwise unkillable town-confirms at times known to the Witches is on inspection one of the worst town ideas of all time.

So yeah, we don't have to worry about Harlot offering degenerate Angel-testing behavior.

Moonlight Pagan

Moonlight Pagan had to change, since it's rules mechanics was built on the old "angel protection" keyword. Plus, no one understood what "inverted" meant.

Of course, Moonlight Pagan is the Irelia of WH. It has been the most pro-town inclusion for ages, and gets nerfed every single major set of changes. This would be no exception. (This was especially true because the main reason Moonlight Pagan had been allowed to be so pro-town in the first place was to make up for Lovers being weak compared to other townfirms. Now that Lovers are getting innately stronger and absorbing more of the town's power of their set, Moonlight Pagan has to be brought in line--worse for town, better for Witches.)

I initially redefined Moonlight Pagan's ability as "That night, the Demons' target is protected instead of the Angels'." I liked this; it had the effect we wanted, but it introduced an asymmetric between Angels and other sources of protection that we had just tried so hard to get away from.

So next I tried "That night, the Demons' veto target is protected. All other protection is canceled." This was alarming, and terrifying! It made the ability SUPER suspicious, while still being extremely pro-town. Exactly what we wanted... but it did introduce a problem. How did this rules text interact with Village Lovers and Harlot's splashing? Technically, it canceled that protection too, which we didn't want.

Thus, those guys got their current "share protection" wording, and we have the end result Moonlight Pagan text:

[Image: 38_Moonlight%20Pagan.jpg]

Note that this allows a hard Witch answer to Lovers. (Though a sacrifice, and use of their kill.) It also "beats" Archangels.

Twilight Pagan

Quick history lesson. Originally, Twilight Pagan was this:

[Image: 19_Twilight%20Pagan.jpg]

This was flavorful, but confusing and had wonky interactions once even/odd night mechanics came into the picture. We moved to the boring but simple version of the same mechanism:

[Image: Twilight%20Pagan.jpg]

Either way, old Twilight Pagan had what I called "hands-tied gameplay." A smart town would tell the TP exactly when they wanted him to go off, when town unambiguously needed a second pro-town lynch. A TP that went off without permission ought to be immediately killed. Even if he wasn't, the lynch kill he had just created was now outside his control. It was a kill that the town had control over instead of you, that you had to reveal yourself (and become Templar-vulnerable for) to use. TP was a definite pro-town influence, but as by far the weakest offensive character in terms of actual player agency.

A predictable result in retrospect: The ability seemed powerful but felt awful.

Also, unlike all other offensive character abilities, TP prolonged the game. Urg.

I considered simply augmenting TP's bonus lynch to only work if he was voting with it, but that didn't solve any of the underlying problems--just alleviated some very surface issues. (In fact, it made the role seem even stronger without feeling or being any stronger!)

Additionally, TP's aw-shucks pro-town gift was... frankly not suspicious? The other Pagans, especially new Moonlight, were crazy sketchy. TP was strictly benevolent at face value.

It was time to make a sketchy Twilight Pagan.

[Image: 39_Twilight%20Pagan.jpg]

It's important to keep in mind a core mechanical and aesthetic premise: survival announcements (like death announcements) can only occur at the very start or very end of day. In other words, Twilight Pagan only works at "dawn or dusk". This keeps the thematic of the role, which is important.

Additionally, like Moonlight Pagan, the core usage pattern for town remains actually unchanged--it's used to repeat-hammer a suspect that survives.

But there are also new town options, albeit high risk, very suspicious ones. Town TP can now also work immediately on King's Court kills, or "call out" a suspected scum self-kill gambit. All this is unilateral and instant. The role fulfills the same purpose, while having much more agency and not prolonging the game.

It also serves as a unique counter to certain dangerous scum strategies. It counters Fanatic just like old TP, but is now also a much stronger counter on Captain. On top of this, it is the only counter to Loose Cannon packing extra lives, with old TP could not counter. All of these events make the TP player look like a true hero. (It's very intentional that TP and Captain are in the same set.)

Witches also now have actual TP gameplay, which is the more important point. Apart from pretending to be a hero, they can make Loose Cannon style sacrifice plays (similar to Moonlight Pagan) in order to take out a high profile target. (Including Lovers, most importantly.)

I'm somewhat concerned that many people will think TP has some sort of infinite loop (it doesn't; it kills during the day, there's nothing that can survive it but finite extra lives), but it's not a bit concern. I might consider rewording the text to something like "They die no matter what." but it's functionally identical and I have mixed feelings. The current text has character.


Templar was a problem for deal issues in Lover and multi-character games. Since the Templar was forced to kill revealed Pagans, it was very unsatisfying to be dealt Templar/Pagan as a single player or pair of Lovers. Also, there was a lingering feeling that Templar was just a worse Wizard, with negligible advantages. So, we're upgrading him in a way that solves the deal issues:

[Image: 41_Templar.jpg]

Now Templars don't have to kill friendly Pagans. They are more of a specialist Wizard than an awkward, bomb-version.


Investigator had two problems. For one, it was slightly weaker than its Inquisitor cousin, because Holy Villagers are just frankly more important to everyone than Village Spies. Second, being an Investigator Spy was lame! Sure, you had the security of knowing that no one (specifically Witches) could find you, and you could play cute games with the Innkeeper, but it still seemed weak, felt weak, and honestly was weak.

We wanted to give Investigator a little bonus that:
  • Was simple.
  • Was generally weak.
  • Was unusually powerful for a Spy Investigator.

[Image: 16_Investigator.jpg]

This is the solution. Finding all the Spies, especially before they (or you) die, is normally quite difficult. Witches could maybe do it, but they would probably have to blow a kill on the Innkeeper first. For Spies themselves though, it's almost trivial. Sure, they have the Innkeeper to worry about, but the option is there, and feels really powerful.

It even seems and feels more powerful than it is. After all, it's just a Survivalist that is conditional on surviving several days and none of your Spy allies dying unexpectedly. However, it seems like you are cheating the system, and you still get the hidden advantages of the role. (Info security, Innkeeper interaction) This bonus makes Investigator way better/cooler at a very low complexity cost.


Old Oracle was lame. It was a simple character, but no one even remembered what it did, because it was so lame:

[Image: Oracle.jpg]

The information seemed powerful but felt weak. It's easy to construct lots of scenarios where knowing how many of X type are alive is very important and useful information, but it normally isn't? The raw census data has no actionable value, in and of itself.

The defensive bonus seems like it was tacked on arbitrarily, which is probably because it was tacked on arbitrarily. Halftime kill roles have to be powerful, because the player has to survive to Halftime. Halftime info roles have to be even more powerful, because the player has to survive through Halftime. Bishop and Zealot definitely fit this bill, but Oracle didn't. So Oracle got this awkward defensive bonus thrown on.

Old Oracle wasn't awful, but the bar had been raised above it.

I spent a long time working on Oracle. I played with a lot of time travel and prophecy mechanics, which are nigh impossible to incorporate into a competitive multiplayer game. One fundamental issue I kept running into was a dual-gated feeling: Suppose Oracle has to make a prediction, and gets a reward if it comes true. Well now Oracle has to both be correct, and survive to see it. The reward has to be pretty crazy, and the role is still doomed to seeming and feeling weaker than it is.

Prediction mechanics also reward/reinforce player group meta, and are particularly unfriendly towards new players who have trouble making important front-loaded decisions. That's a dead end.

The other class of mechanisms I was considering was some sort of Halftime Innkeeper; someone who could watch the Halftime roles. They could retain the Halftime immunity, and it would now make sense and feel appropriate. We knew from the get-go that they couldn't be allowed to watch the Bishop, Spy Saves, or Angel/Demon Saves. It also didn't take long to conclude that Oracle would be virtually strictly superior to the Zealot if she could watch him, as well as ruining his powerful and fun pro-witch gameplay. That left the killers: Hunter, Entertainer, Assassin, and Summoner.

However, Halftime Innkeeper had a slew of universal challenges across all implementations. The timing was a nightmare for online games, because suddenly all Halftime kills had to be completely sequential. The killer had to be informed, they had to have time to discuss, and input a decision. Then the next killer, in sequence. And now, for the first time, order mattered. It was a complete mess and there was no true way around it.

What if Oracle found out their identities and targets after the fact, unilaterally? She still needed protection to work reasonably, which was suddenly weird and unsatisfying. The interaction was lost, and the role was actually really weak--it was just some low-value town accountability, with trivial value to witches. What's the point of knowing they they killed, when the bodies are already dead?

That's how we finally arrived at the solution:

[Image: 23_Oracle.jpg]

What if the Oracle found out their identities before the fact? Say, at game start? This "prophecy" is interesting because it seems and feels powerful, and is powerful, but for different reasons. Knowing arbitrary neutral roles isn't actually that potent in and of itself, but knowing 4 non-Holy-players from the start of the game sure is; that's a massive witch asset!

We even list the Halftime kill roles in the flavor text to kill the FAQ.

New Oracle is a really unique and cool position in the game, with a really simple mechanic that adds some needed cohesion to the Halftime expansion. I'm really excited to see people play with it.


Old Paladin gave an extra life at Halftime. This was okay. He worked fine mechanically.

[Image: Paladin.jpg]

What's the issue? Paladin was in the King's Court expansion, not the Halftime expansion. There are no Halftime cards in the other sets, just as there are no King's Court cards in the other sets.

Does this matter? Yes! Set organization is a critical weapon for controlling complexity, and providing cohesive experiences. They help give games stronger narratives and identities, and optimize the rules burden. It wasn't a big deal, but there was definite motivation to find an alternate trigger that was just as good but didn't involve another set's mechanic.

There was also a slight concern that a Witch Paladin would find themselves in a "hands-tied" situation, where it's halftime and they have no choice but to give a townfirmed on the court (usually the King) the extra life, or be exposed. Sure, a Witch Paladin should have been able to kill off the townfirms on the Court to prevent this, especially the King, but it's still a bad experience (with no gameplay) that we'd prefer to prevent.

The solution?

[Image: 27_Paladin.jpg]

By using the Court itself as a trigger, the Paladin gains a much stronger identity. He's the last line of defense, defender of the court!

The new trigger is significantly more rare; it's about a ~33% chance of actually triggering, much lower than the old ~50%. This makes the Paladin one of the rarest-to-trigger roles, on par with the Fanatic, Anarchist, and Wizard. (However, keep in mind that the Paladin is an automatic Court member regardless.) The rarity of his new trigger has two important ramifications.

First, the Paladin no longer has to be a one-shot. This seems and feels super powerful, but since he only gets to use it every other night, the power is bound and counter-able.

Second, Witch Paladins have much more time and freedom to eliminate townconfirmed players on the Court. It is far harder for this new Paladin to find himself in a "hands-tied" gameplay situation.

New Paladin is downright cool, and adds to the King's Court subsystem in multiple ways.

Groundskeeper (Removed)

[Image: Groundskeeper.jpg]

We talked a bit about Groundskeeper in the other thread. The idea of a neutral role that knows a (weak) Holy is interesting, but potentially dangerous. I'm not going to repeat that entire discussion here, but the bottom line is this: If a Deacon gets shot night 1, and an Elder was the Groundskeeper, they are going to be furious and feel like there was nothing they could have done. Even if that's not true, and even if sniping the Deacon is a dubious decision, the miserable experience that Deacon will feel is an non-debatable fact.

On top of that, Groundskeeper isn't even that cool. Acolyte has always been more cool and exciting than Nun, but Nun is still interesting and both enjoy all the fun gameplay perks of being Holy. Groundskeeper is lamer than both and enjoys none of that. It's a very inert role that seems like filler. It's easily forgettable, and no one is especially fond of it.

The bar has been raised past Groundskeeper. We're cutting it loose to make room for something way better.


So Traveling Mortician has been a huge success:

[Image: Traveling%20Mortician.jpg]

Night death info is a really cool mechanic that I've always wanted to pursue, but is unintuitively awkward to implement because of the timings of death resolution and announcement. Travelers provided us with a night 3 trigger, that was the perfect time to weasel in a night-death info mechanism. Yeah, it was kinda awkward, but I predicted it would be worth it and I was right. Traveling Mortician has been one of the most successful and well-liked roles by players.

Quick analysis aside: night-death info is more powerful (and more interesting) than info from lynches, like the Gravedigger gets. This is because townfirmed/townfirmable players die almost exclusively at night, and this is outrageously valuable info to Witches in particular. For example, Witches that know they have killed the Priest can claim Priest.

Traveling Mortician wasn't perfect though.
  • He stuck out in the Travelers cycle like a sore thumb. The other Travelers all select a single target, he selects no one. They all see/share each other's information, he keeps his info private. He has no interaction with them at all.
  • The other Travelers knowing that he was the Mortician lowered the value of getting the information. It also "half-tied" his hands as a Witch; he would be expected to come forward the next day as the Mortician and say something.
  • Meanwhile, the Mortician being privy to other Travelers was excess power that wasn't relevant to his identity or core gameplay.
  • While there are normally 3 dead bodies for him to look at (N1 Coven, N2 Coven, N2 Court) some may have survived, lessening the power of his role in a lame way. It's even possible that none died, screwing him over.
  • I was considering a new Traveling Minstrel, who the Mortician didn't work/interact with.

The Mortician needed to move out of the Travelers. He had earned his place and we absolutely knew we wanted to keep his nature of mechanic. I didn't want to keep the name Mortician, because I was never that fond of it and players might find it confusing. (Non-Traveling Magician is bad enough, though it hasn't been a problem since Magician is so memorable.) We still had no character that started with U--this was perfect.

[Image: 29_Undertaker.jpg]

(hey look, placeholder art from Groundskeeper--HE'S DIGGING UP DEAD THINGS ALREADY)

So Undertaker's core mechanic is waaaay more powerful. He's not accountable to anyone, and he can get more info later if he is willing to wait. (Or less sooner!)

The first way we restrain his power is by giving him unlabeled sets of cards; he doesn't know which cards belonged to who. This puts him in the fascinating dilemma where the longer he waits, the more info he gets but the less certainty/clarity there is about any of it. It's a really cool question to consider!

This decision point creates a cost, particularly for offline games: we have to call for Undertaker every night. Undertaker seems great, but is he worth *that* cost? An Undertaker also needs a potentially long time to review the massive pile of cards, so this is an even bigger cost than any other night action. (Compare to Harlot, who is relatively light.)

On top of that, the role is still too powerful, mostly on the Witch side. But the two problems share a common solution.

We publicly announce when the Undertaker activates. This adds some very light accountability to the role, just the right amount of restriction on Witch Undertakers that we want. And suddenly in offline games, Undertaker only has to be given time for a thumbs-up/down each night; we only have to give the full time when the ability is announced, and then we never have to call for the role again! The night time cost is suddenly much, much lower, to something we are willing to pay for such a great character.


[Image: 28_Vigilante.jpg]

Speaking of time costs, we have previously been announcing Vigilante triggers publicly, at the start of the night he kills. Another way of saying this is that we implicitly reveal when he does not trigger by not calling for him. This saves us a bit of time.

But does it really? The Vigilante only has to be called on odd nights, starting on 3, until the Court dissolves. He is generally called for only 2-4 times per game. Is there really that much to be gained by keeping the trigger public to avoid this cost? What do we gain by having it non-public? Well:
  • Other night actions do not get awkward advance/parallel info about the trigger.
  • If Night 3 only has 2 kills, the Vigilante's can be mistaken for the Mercenary's.
  • Witch Vigilantes can elect not to kill, to fake no-trigger for whatever reason.

The first two are mostly insignificant, slightly pro-witch implications. The third is a pretty cool, non-trivial pro-witch option that adds some depth and gameplay to the role. All of this is desirable, because as we are about to see Travelers are taking are hard shift in the pro-town direction. (Undertaker being a considerably pro-witch inclusion helps offset this, but a tad more is welcome.)

Anyway, I am willing to pay this cost. Vigilante no longer has public messaging.


I've already spoiled that the old Traveling Minstrel is getting removed. Why?

[Image: Traveling%20Minstrel.jpg]
  • 9-player Courts were proving to be too big.
  • The role was downright weak. A Court promotion isn't that big of a deal...
  • ...nor is it in his control. The Witch Minstrel was the ultimate example of "hand-tied" non-gameplay. He had no choice but to put who the town requested on the Court, or else he would be exposed as a Witch. His choice wasn't actually his choice!
  • It's not satisfying to put someone else on the Court. It just... isn't. Picking someone else to have fun is lame.
  • The Minstrel got no feedback--he had no way of knowing if he added someone successfully, or if they were already on the Court and he did nothing. Was it a good choice? Was it a mistake? He has no clue, nor any resolution of any kind.

So I already knew that we needed a new Minstrel. I really liked the idea of a Minstrel, but what could he mechanically do? In traditional fantasy, this class of character tends to support their teammate's actions. Could I do something like that?

[Image: Bard.jpg]

Oh, hell no.

All the way back in WH 5.0, in December of 2010, I made a character called the Bard. He could "amp" one target a night, and every single town role (this was pre-bifurcation) had an amp ability. Every single character had a second effect written on the card, solely to support this one character. It was a disastrous experiment that doubled the game's complexity for paltry benefit.

You can't even do something intuitive and implicit like "Select a target, double their power." That's very ill-defined; what on earth does that even mean to half the character abilities? Double Lookout? Double Innkeeper? Even something as seemingly straightforward as Double Survivalist is nebulous: Does he get another extra life? What if he already lost it? What if you already doubled him, does he now have 4 extra lives? What if he got an extra life from somewhere else? Technically his ability is that he gets an extra life at game start, so since it's no longer game start, shouldn't it have no effect on him? This sort of path is a disaster. (Along similar logic, we can't have mimics.)

Any sort of implicit amp is doomed to only work on a small subset of roles. And how on earth is the bard supposed to know who he can and can't amp? That's impossible, he would have to be in some sort of group with them.


[Image: 36_Traveling%20Minstrel.jpg]

New Minstrel is an amp role that only applies to his fellow Travelers--players he knows. He implicitly doubles their ability, letting them pick a second (different) target.

A Witch Minstrel should never have his hands-tied. None of the others should be forced to claim/plan publicly, and all are viable options for him to consider. (Even when the other 3 are all town, he can discuss their plans and pick the one he approves of the most.)

[Image: 33_Traveling%20Medic.jpg]

The Medic is losing his auto-notification. That was arguably too pro-town before, and would definitely be way too pro-town with the new Minstrel; due to the way the value of that info "stacked", not only was this too powerful, but it tied the Minstrels hands into always having to amp Medic as the definitive pro-town choice. It risked a lot of "hands-tied" gameplay for other Travelers as Witches, which we also want to avoid.

Now the successful Medic target(s) receive notification privately, after the Travelers have concluded. This means they can still step forward and confirm the alignment link, or they can lie! It's still a super pro-town inclusion, but it's bound and counter-able now, and a safe thing to allow doubling.

[Image: 34_Traveling%20Mercenary.jpg]

Traveling Mercenary is unchanged, but changes to his fellow Travelers redefine some of his dynamics. He is less beholden to assassinating people that fail Medic targets, and is a solid Minstrel option.

So we obviously needed a new Traveler, to fill the Mortician's vacated spot. The new guy needed to be something that selected a single target, could be doubled, and didn't create hands-tied Witch gameplay for himself nor other Travelers.

I spent a lot of time trying to generate functional info-role Travelers. This was a fruitless path for a few reasons:
  • It's tricky to have an info role in the Travelers, a group that sees each other's stuff. We didn't really want to repeat the awkward Mortician example, where one of them just doesn't share.
  • It's tricky to have powerful one-shot info, something that is actionable and meaningful. By nature of mafia, most powerful info is either very complicated or degenerate in consequence.
  • It's tricky to have a powerful info role that works by targeting a single player. How much (non-degenerate) info can you really extract from one guy? Can you really make a compelling one-shot?

I wanted to name the role (that didn't exist yet) Medic, Medium, or Merchant. I considered a lot of "bribe" mechanisms for a Merchant, but they were all awful; Merchant implicitly calls for resource systems, a massive complexity investment that was never ever happening.

I considered a Medium who could check if a target was picked by the Angels or Demons, and got some reward. This was awkward and weird, really bad for novice town, and there was no appropriate reward. However, this led to:

[Image: 35_Traveling%20Mystic.jpg]

Dead player interaction roles are super hard to design. And yet this one fell into my lap while I was hell-bent on crafting an info role. (That could probably never actually exist) It gets away with this by interacting with dead players before they are dead.

This is obviously a pro-town inclusion, like the other Travelers, because Archangel cards are so much more powerful than Archdemon cards. It's hard to estimate, but I suspect the power level is about right.

An important dynamic that makes the character non-degenerate is the value in keeping the selected target secret, if possible. Witches would like to know about the trap that has been set, so they can consider avoiding it (at least until they have nuked important targets like the Priest) or if nothing else be aware when an extra Archangel is on the table. Ideally, a town Mystic traps a Villager they believe is going to be killed by the Witches ASAP.

We avoid telling players they have been selected by the Mystic, becasue it doesn't need to be more powerful and players should never feel uniquely compelled to die. (Lest the card "go to waste")

The new Traveler cycle is cohesive, mechanically sound, and has a lot of narrative/identity going on. New Minstrel in particular is one of the changes I am most pleased with.


Captain has been a concern in multi-character games. It stacks way to well with hard defenses. Even if rules changes about Magician have killed that interaction, Pauper and Survivalist are still issues. Finally, even in non-multi-character games, allied B.O.D. or Medic makes a terrifying threat. So, we're nerfing Captain in the multi-life case:

[Image: 44_Captain.jpg]

So now if you lynch the Captain and they survive, they don't get the lives. In 99% of cases, nothing changes. The Captain does gain one perk from this: If the town lynches the D.O.B. instead of them, they get the lives immediately and survive any D.O.B. kill thrown their way. (Leeroy and Judge still take them out, of course.)


There were the original three Musketeers:

[Image: Hasty%20Musketeer.jpg][Image: Brave%20Musketeer.jpg][Image: Sassy%20Musketeer.jpg]

There were three big concerns, appropriately enough.
  • The Musketeers were a little too strong, mostly in the Witch direction.
  • The Brave Musketeer was really awful and weak for town. (Largely the same as Hasty for Witches)
  • Lots of players had trouble remembering what they did, and if they worked on team vs. type.

I made a lot of Musketeer designs. A lot. Most were bad, or at least had one bad apple that ruined the bunch. (Sort of like Brave!) There was even a "vengeful" cycle of Musketeers that I was convinced was perfect, until I found minor issues with two of their possible team assignments. I'm not going to cover all the proposals considered, as that would take forever.

Ultimately, I knew I wanted the two following "safe" Musketeers:

[Image: 46_Pious%20Musketeer.jpg][Image: 47_Sneaky%20Musketeer.jpg]

These were a little better for town (but felt quite a bit better!), and a moderate amount worse for Witches in an interesting, appropriate way.

Why not do a third one for Village Lovers? Much of the Musketeers' power comes from their ability to kill Lovers; a Musketeer that couldn't kill Lovers would be a definite runt. Town has relatively little need to ever prove or confirm Lovers, and Witches value killing them as a unique opportunity.

I was talking to Tea about it, and he suggested something really stupid. I told him how stupid it was, then immediately implemented it and never looked back.

[Image: 48_Reckless%20Musketeer.jpg]

At face value, Reckless Musketeer seems terrible for town. It can't kill Witches! But ultimately, that's kinda the point.

Take a step back for a second, and consider a basic town kill. There are two possible outcomes:
  • It hits a Witch, and the Witch dies.
  • It hits a Villager, and the Villager dies.

Now compare to a kill that doesn't fire on Witches:
  • It "hits" a Witch, and they are publicly hard-confirmed as a Witch. (but not yet dead)
  • It hits a Villager, and the Villager dies but is publicly confirmed to be a Villager.

The first case "Reckless kill" has pros and cons: It's worse closer to "mylo" when you just want to shoot and pray, but is way better in the endgame where you still have an abundance of town kills and a shortage of hard information. (Hard confirming a witch who has been claiming an info role can be a devestating blow!)

And for the second case, though it might feel bad, the "Reckless kill" is clearly strictly superior for town!

(Side note: hard-confirming Witches is only permissible gameplay because we have been vigilant about keeping out any mechanical motivation to keep a known Witch around. It's almost impossible to construct scenarios where town doesn't immediately lynch the Witch.)

Meanwhile, Witches get to kill whoever they want, at the non-trivial cost of retrospectively confirming the targets are town. Which in some cases, like going all in to blow up two pairs of Lovers, is already implicit. It's an appropriate restriction.

Multiple people have wanted Reckless to still kill Juniors, but that sort of undermines the point. The value of Reckless Musketeer's kill is hard info, with no exceptions.

For awhile I changed these Musketeers to "Kill two targets. Any ____ survive." such that survivals from extra lives were indistinguishable. This had somewhat interesting gameplay with Pious and Sneaky, but undermined Reckless. I changed it back.

I also changed the trigger to be free floating, and not strictly reveal-on-second-Musketeer's-death-or-lose-it. This feels better, is stronger, and has cleaner interaction with Magician in multi-role games--but also has some downsides. I like that the old trigger allowed Lover Musketeers to exist, and I liked that it seemed to discourage Witches from sniping other Musketeers early. Reverting this perhaps the only mechanical change to anything I am still considering, pending feedback and data.

Flavor Text

I might write a bit about the changes in flavor text paradigm later. Long story short: It turns out flavor text is a critical tool in establishing perception of understanding, and I had been ignorantly overlooking this for years.

What Now?

Well, now I go code this stuff up. I've already laid most of it out, just gotta do the nitty gritty. Shouldn't take more than 3 days; Summoner is the only big task coding-wise.

So... see you soon?

Edit: Here's a link to drafts of all the card images and a rules PDF.
04-21-2013, 09:59 AM
Post: #2
RE: Design Review: 9.0 Design Changelog
There are no cards for summoner and investigator, just some bolded text that says "summoner image" and "investigator image". Since the text assumes that you've read the cards I'm not quite sure what those roles actually do.
04-21-2013, 02:48 PM
Post: #3
RE: Design Review: 9.0 Design Changelog
I wholeheartedly approve of these changes and your methodology.

I always felt that Oracle was a really weak role and I think that change is great.

And kudos to Tea for suggesting the reckless musketeer because that's a super interesting play mechanic.

Oh and totally agreed about flavor text. Intuitive understanding from the names and contextual understanding from the flavor allow a game like this to suffer the complexity required to manage this many players each with unique roles.

Damn fine work, sir.

Doctor Who: The Forgotten Doctor
Escape the Day
04-21-2013, 06:43 PM
Post: #4
RE: Design Review: 9.0 Design Changelog
These are some pretty rad revamps, can't wait to give it a shot. Seconding Scoffles re: the flavor text, it's often under-appreciated for how much more understandable it can make the game when worded correctly. Not to mention the fun/narrative aspect of it.
04-22-2013, 12:01 AM
Post: #5
RE: Design Review: 9.0 Design Changelog
(04-21-2013 02:48 PM)Scoffles Wrote:  And kudos to Tea for suggesting the reckless musketeer because that's a super interesting play mechanic.

I throw out random ideas to see what sticks. Sometimes you get the virus. Sometimes to get Reckless Musketeers.

Awesome work, CPi! Now host a game already!

Forum Jump: