The One Time I was wrong while designing Secret Hitler

For the last year I’ve been working on Secret Hitler, a game of hidden identity where everyone claims to be on the Liberal team, but some players are secretly on the Fascist team. It features an Investigation mechanic where a player gets the power to find out for sure what team another player is on, which is extremely valuable information, but the information comes at a cost that makes it harder to share.

Another aspect of Secret Hitler is that one Fascist player’s secret role is Hitler. If the Liberals assassinate Hitler, they win the game. But if the Fascists get Hitler elected at the right moment, the Liberals lose immediately. The drama this creates is part of what makes Secret Hitler so fun.

Kill. This. Lizard.

In early playtests, it was possible for the Liberals to discover Hitler’s identity with an investigation. We found that when that happened, it sucked all the dramatic tension out of the game. There was almost nothing the Fascist team could do to recover. They just had to wait for the second half to roll around so that Liberals could use the assassination for the easy win.

In every other instance, both investigations and assassinations ramp the tension up to 11. But in this one specific intersection, when the dramatic heart of the game gets uncovered too early, they weren’t fun at all. We had to devise a way for players to find out what team someone was on without knowing their specific role, all in order to prevent Hitler from being found out too easily.

What seemed like an easy design problem, though, led to hours of debate over which design we should use and why.

Here’s where we ended up: every player gets both a Secret Role card that designates them Liberal, Fascist, or Hitler; and a Party Membership card that designates them as either a Liberal or a Fascist. They never get mixed-and-matched; it’s not like Liberals ever have a Fascist Party Membership or vice versa. A player’s Party Membership card always contains the exact same information as the Secret Role card, except when a player is Hitler. That’s the only time it really matters.

Redundancy everywhere.

Literally 9 out of 10 times this design is completely redundant, which is why I hated it to begin with. But I think the story of how we ended up there is really interesting, and I learned a lot about component design along the way.

Design Alternatives

Fans who reach out to us about the Party Membership card tend to have two complaints:

  • Ten totally new components seems like an unnecessary number of components to solve this one edge case
  • Because the two cards are separate but usually redundant, new players don’t understand their purpose and worry that a Fascist might get a Liberal Party membership card, or vice versa.

Along the way we prototyped two other designs, both of which used far fewer components and were more readily understood by new players — they lacked the problems of our existing design.

The first design, before we even had graphics, had Reveal cards that said either “I’m a Fascist!!” or “I’m a Liberal!!” The President investigated a player by handing over these two cards. That player would awkwardly shuffle the cards, pick the right one, and slide it to the President to check. Then the President would slide it back and the player would awkwardly shuffle them again and set them aside.

Mike wanted the left one to say “I told you I’m a Liberal!” but worried that Presidents might laugh and give too much away.

The second version was a single unified Secret Role card that had the team information at the very top. When investigated, the player would slide the card part-way out of an envelope and then show that to the President.

The top band only reveals Hitler’s team, not his role.

Confession time: when we started, I would have preferred either one of those designs to the one we settled on. I liked the simplicity of the two Reveal cards and the extreme parsimony of the banded Role cards. Fortunately, my co-creators fought to give them a thorough vetting, and through playtesting, though, we discovered less obvious but far more important problems with each one. The rejected designs share two problems that our final design solves:

  • Bad emotional fit: both designs felt awkward or weird or anxiety-inducing-at-the-wrong-time to interact with, which hampered people’s play experience.
  • Catastrophic failure: Not, like, an earthquake or anything. No design totally prevented human error, but our rejected designs failed in ways that ruined the game and, more importantly, were hard to detect, which means that games that had already been ruined would carry on for like an hour before players realized their time had been totally wasted.

Emotional Fit

Secret Hitler encourages players to be extremely paranoid about accidentally disclosing their hidden identity. Controlling the flow of information is what the whole game is about. What I didn’t sufficiently appreciate is the way that paranoia leaks into player interaction with the components.

With the two-card design, players would shuffle the two cards a bunch before sliding the right one over, and then shuffle them again before setting them aside. They were extremely anxious that an attentive player could, just by watching their hands, discern what team they were on. It also made players feel, subconsciously, like the information of what team they were on was still physically on the table, in the form of whichever card they had passed over. They had no way to protect that card from further scrutiny. Extremely unpleasant.

If you want a taste of what that felt like, try this: take two index cards. Put a black X on one of them, turn them both over, and while someone is watching, shuffle them back and forth on the table. At the end, you want to still be able to identify the one with the X on it without looking, but you want the person watching to have no idea which is which. It’s tougher than it sounds! And it led players to dread interacting with the two Reveal cards. They worried they would be revealed because they lacked dexterity, which is not really a skill anyone thinks is core to the game experience.

We also consistently find that players are very paranoid about protecting the contents of their envelope. Typically when envelopes are dealt, players peek inside to see which team they’re on without ever removing their cards at all.

Hard to tell, but this is actually the Hitler card.

With the top-band design, every couple games or so a player would see “Fascist Team” at the top of their card and not realize it said “Hitler” a little further down. It happened even to very experienced players. We tried reminding players to look all the way down into their envelopes, but that didn’t really work! Besides, constant reminders isn’t a sustainable design solution.

(This is also one reason we abandoned the dossiers, which you can see over in Max’s prototyping post: there was a ton of helpful information in there, but players never opened their dossiers after the initial deal, because they didn’t want to expose that information.)

So, these designs created conflict with the emotional theme of the game, which is Bad.


This version is much harder to get wrong.

The solution that we settled on, a separate Party Membership card in every envelope, feels much better to interact with. Players can now see their identity without removing anything from the envelopes, and the card they need to show the President remains safely in the envelope until it’s needed.

After an Investigation, the card can be returned safely to the envelope without any fear of someone tracking your hand movements. Players are then free to tuck their envelopes, and their Party Membership with it, into a back pocket or under a thigh or wherever they’re safely keeping it. Players only feel exposed at the exact moment they’re meant to.


Failing, good and bad

Since no design can totally eliminate human error, a good design approach should:

  • make things as easy as possible to get right
  • make it easy to detect when something has gone wrong
  • minimize the impact of errors.

While our rejected designs might feel more parsimonious or intuitive that the final design, neither one failed particularly elegantly.

I’ve already talked about the tremendous anxiety the two-card design causes players about getting it right — making sure they get the right card without anyone else finding out. If a player does manage to track the awkward shuffle successfully, that player has an unfair advantage for the rest of the game. But there are a couple other ways things can go wrong, and they’re both pretty bad.

First, they might slide the wrong card over and no one notices. In that instance, the game continues on for another hour and bitter, angry arguments are had because two Liberals are each convinced the other is a Fascist. At the end of the game, when everything is revealed, those Liberals will be pissed that a drag-out game was caused by a preventable card error. Alternatively, someone might actually notice (say, a Fascist who has perfect information) and say something. At this point, the investigatee’s team has been revealed to the table, and the person who spoke up has probably also outed himself.

Either way, the game is ruined when the two-card solution fails.

The top-band design failed even more frequently than the two-card design. One thing that kept happening during investigations was the President would accidentally slide the card too far out and see the body color of the card, which would tell him a player’s role as well as his team.

Just a little too much information, there.

We could fix that problem by making the band thicker, but that made it harder for players to see their role just by peeking in the envelope. With the thicker band, we would get to the part of the game where the fascists learn each other’s identity and find out who Hitler is, and no one would identify themselves as Hitler. That forced us to start the game over and give everyone a new role.

These failures weren’t as catastrophic as the two-card design’s failures, but there was no way to detect these failures before they affected the game. Anytime someone got too much information, there was much hand-wringing about how to proceed without starting over, because no one wanted to junk the game completely. When we did have to start over, players who were excited to play whatever role they received (“aw, man, I finally got to be a Fascist!”) would inevitably feel cheated of an exciting experience. It was just frustrating.

In contrast, a separate Party Membership card given to each player fails much less often, and more elegantly.

It’s easy to get right, because you only have to pair each Role card with a Membership card once, during the initial setup. After that, it’s just “pull out the yellow card and slide it over.” No fussing with extra cards, no pulling it out just the right amount like some kind of Hitler Goldilocks.

It’s easy to detect failure, because the Role cards and the Membership cards have distinct designs. Usually the table corrects a player who pulls out the wrong card by saying, “hey, not that one, the yellow one.”

We can tell if someone has pulled the wrong card before any damage is done.

And when that failure is detected, it’s not catastrophic — the player can return their Role card and safely pull out the Membership card with no negative consequences.

What are we designing for?

For me, this process was a really valuable lesson in design tradeoffs: what do we want to prioritize with our design choices?

We sometimes get emails proposing variations on the design that require fewer cards or are easier for first-time players to understand. Typically they’re variations of the two specific designs we tried already. Those designs prioritize parsimony of components, and there is something really, really satisfying about a solution that accomplishes a bunch of goals with a minimum of bits and pieces. That’s why I initially preferred those types of solutions, and I imagine it’s why so many players bother to reach out to us with alternatives: they scratched an itch in a satisfying way and they want to share that with us.

Those alternative proposals are also frequently motivated by a desire to simplify explanations. Players always have questions when they look in their envelopes and see two totally redundant cards, and sometimes they’re afraid to ask for fear it will reveal what team they’re on, and it means we have to explain a rule ahead of time, even though it doesn’t matter immediately. In other words, the design we’ve chosen is less beginner-friendly than the alternatives.

On the other hand, once players do understand this design, it’s the one they have to think or worry least about. It’s the hardest design to mess up and the one most in keeping the emotional theme of the game. While it’s less beginner-friendly, it’s most veteran-friendly. It’s the solution that makes the game the most fun to play over and over again. And we’ve never found that players struggle for particularly long with the separate-card design; usually once they see it in action at the first Investigation, they’re on board.

Most people play Secret Hitler again.

As of March 24th, 2016, we had 578 responses to our playtest feedback form, including three responses from March 23rd. 522 of those responses (a little over 90%) indicated that the group had played at least two games; 267 of them (46%) reported five games, which is the most our form allows reporting.

All of which is to say: most Secret Hitler games are played by people who have already played before.

That’s great news for us, because that’s what we’ve prioritized: our design has a small upfront cost that might be too high if no one bothered to play Secret Hitler more than once. It’s aimed at making the long-term experience of playing Secret Hitler both emotionally congruent and easy to get consistently right.

Never Don’t Playtest.

Any of these three designs would have functioned perfectly fine within the rules, but two of them made play frustrating and easy to mess up, which would have made it less likely for people to share with their friends or even bother to play a second time.

Most importantly, this is something we only could have worked out with playtesting. Playtesting isn’t just about getting the rules right; it’s about getting the entire experience right. After all, games are created for players, not players for games. When players didn’t enjoy the design I thought was best, it was an easy choice to lose the design rather than the players.

I’ve written about designing Secret Hitler before! Learn about how we designed the hidden information mechanics and how we balanced the policy deck to make the game exciting. I also talked about the Liberal vs. Fascist balance here. And for more on the value of prototyping, Max Temkin wrote about the process of prototyping Secret Hitler more generally and has some extremely cool images to share.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.