Brooke Ballhaus, Jeffery Cai, Lily (Xinrui) Li, Sabrina Yen-Ko

Artists’ Statement

The earth has flooded! The mammals have built three arks in order to survive, but the reptiles–each of them avid swimmers–have other plans: to sink the arks and dominate the planet.

The Golden Ark is a strategic social deduction game, where mammals have to deduce who’s on their team and safe to board an ark, while reptiles try to evade suspicion and slither their way onboard. The reptiles sink the arks, but two heavy mammals will break one just as well! Two down, and the mammals are kaput. Through psychoanalyzing your animal friends, strategic resource allocation, and hopefully some mistakes by your opposition, players will sail (or sink) towards victory.

This game is intended for engaged players who love complex gameplay and strategy. With more animal identities than players and dubious investigations, absolutely nothing is certain. And if you’ve spent the whole game saving your gold, your friends may reward you with a coup and take it all away.

Dive into this magical world with friends or with strangers. Either way, you’ll learn plenty about your fellow animals’ strategic abilities and acting skills (or lack there of).

Get ready to know nothing, trust liars, and get your paws damp.

Concept Map

Initial Decisions

When creating this game, we all had one goal in mind: making a game that we found fun. For Lily, this meant having unique artwork (preferably with animals); for Jeffery, this meant all players should have a unique role; for Sabrina, this was about player agency; for Brooke, this was about promoting deep analysis and deduction. Our target audience was decidedly people like us: lovers of social deduction games who are comfortable with social analysis, complex gameplay, and strategy. As social deduction lovers, we had a few gripes with many of the games we had played. We were able to agree on a few tenets, based on our individual definitions of fun and our qualms with other social deduction games:

All players stay involved throughout the game (no killing)
Every player has an active and somewhat balanced role in the gameplay
The game necessitates deep engagement in order to solve
Cute animals

From this point, we began to think about the aesthetics and dynamics we most cared about so that we could create mechanics that promote them.

Fun as Fellowship: We wanted to emphasize the importance of both teamwork and competition between teams, especially the need to bluff to your opposition. We were particularly interested in creating dynamics of uncertain alliances, such that players have to put their trust in one another without knowing whether their objectives are aligned. To promote these dynamics, we started with the following mechanics:

The Reptiles know each other, but the Mammals do not – allows the Reptiles to have a teammate to rely on and requires that they then must fool the other team as to their alignment
All players can investigate one another – allows for Mammals to create alliances based off of uncertain information

Fun as Discovery: One dynamic we enjoy is that of discovering new strategies and utilizing them. As well, we enjoy the dynamic of having to “solve” a game. For these dynamics, we started with:

Mammals have to figure out who is on their team in order to win – requires the Mammals to solve the puzzle exactly of who is who
Every player has a role with abilities that add complexity to the game – enhances the complexity of gameplay, as well as the possible strategies that can be utilized (e.g. the snake discovering that they needed to silence the rabbit to win the final ark)

Fun as Challenge: We wanted the game to be replayable and to feel difficult. To do this, all players needed to be engaged and equally a part of the gameplay. As well, we wanted dynamics where winning was never easy for either team, thus the game had to be balanced. The game also had to be complex so that there is never one optimal strategy. To create these dynamics, we introduced the following mechanics:

Every Mammal role was somewhat unhelpful to the Mammal team – helps to balance that the game is four mammals versus two reptiles (e.g. two heavy animals that could sink arks)
Resources must be managed for both voting power and information gathering – requires players to think about possible future scenarios and allows for infinitive iterations of the game

Fun as Narrative: We liked the idea of having a consistent theme and narrative present in the game that allowed players to engage more deeply. Thus, we had the following story:

Earth has been flooded! There are two distinct classes, the mammals and the reptiles, who are both trying to get on a final ark, where the animals on which will repopulate the earth. There are two test arks, and one final large ark. The reptiles, as they are avid swimmers, plan to sink the arks to keep the mammals from repopulating the earth.

Fun as Fantasy: We wanted the experience of each player to be immersive and fun, so we included a light-hearted narrative. We also wanted players to have fun by playing as their characters or making light of the animal characters. So we implemented the following mechanics:

Each player has a unique role – allows the player to take on a new identity
Animal abilities have some relation to the animal’s real-life qualities – provides depth to the fantasy

Fun as Sensation: We all agreed that one of the best part of board games is the physical pieces that players can play with, so we knew that we would be creating artful objects and mechanics to allow players to experience fun physical sensations, such as:

Using gold nuggets with texture as votes
Having lasercut tokens and bags for the ark expedition voting and results
Using a passport to hold player identities
Having a consistent theme for artwork

We also took a few mechanics from other social deduction games, such as the “quests” and nominations from Avalon and the night phase from One-Night Ultimate Werewolf. Our initial rulebook in its entirety can be seen here: CS247G P1 Brainstorming Ideas.

Initial Formal Elements

Objective:
The mammals want at least two boats to succeed (must find out everyone’s alignment – solution).
The reptiles want at least two ships to sink (must hide their identities – outwit).

Players:
The game is for 8 players, five as Mammals and three as Reptiles.

Procedures:
Starting actions include game setup and reptiles learning who each other are. The core loop proceeds as follows:

Discussion phase — players investigate and discuss.
Nomination phase — one player nominates three/four players to go on the ark; if the vote passes, the ark sails.
Expedition phase — the selected players each submit either a mammal, heavy mammal, or reptile token in a bag.

The loop continues until two arks have either sailed or sunk, or three rounds have passed (whichever comes first).

Resources:
Each player starts the game with 11 gold, which is used to pay for investigations and to cast votes.

Conflict:
Mammals don’t know anyone’s alignment. Players must manage resources to gain information and voting power.
Reptiles are outnumbered.

Outcome:
Mammals win if two arks successfully sail (or if one ark sails and the rabbit is on the final boat).
Reptiles win if two arks sink. The outcome is zero-sum between the teams.

Iteration History

Playtest 1 (Thursday during 3B – 30 min):

We had the first playtest during studio 3B with six members from the class. This session was rather short, and players were only able to get through the game set-up and the first round. The key feedback was that – though the players loved the theme, the animals, and the artwork – the game felt too much like Avalon. The party leader chose an expedition party, but the rest of the group didn’t get to do much, as the point of investigations wasn’t clear. Neither was the secret voting with gold. This mechanic took too long and was disengaging to players. We knew we wanted to keep the artwork and theming, but change the mechanics.

Iteration – Gold distribution as party nomination

Allowing players to vote on the expedition party with gold did not create the agency we wanted, added downtime where players could not make actions, and felt too similar to Avalon. We fixed all three of these issues by allowing all players to independently and publically decide who to fund to board the ark with their gold. Every player now plays a mandatory and important role in choosing the people who determine the success or failure of the ark and have a more equal influence on the exact players, rather than just voting on a party someone else chose. Additionally, public funding encourages players to remain engaged in the funding process and other players’ decisions are immediately obvious and may affect their own. Finally, this change was a much larger deviation from Avalon.

Playtest 2 (Sunday afternoon in CoDA – full game):

For this playtest, we played with a group of 6 volunteers and made the following observations:

In the funding phase, the last player to distribute gold was able to unfairly influence the auction, allowing a reptile to automatically board the ship and waste all mammals’ gold on that ship.
- This did not feel like a fun dynamic, because the player who begins the funding phase was largely chosen at random, so mammals did not feel like they had agency in stopping this potential outcome.
Reptiles are unable to successfully bluff as the Beaver or Cat because those roles could choose to reveal for a benefit.
Mammals knew exactly which reptiles were in the game, but reptiles did not.
- This allowed mammals to always be on the lookout for these roles, making it difficult to bluff as a reptile.

Iteration – The Auction and New Roles

We added an auction phase to counter the funding phase being based on lucky seating arrangements. Players now bid for both the maximum amount of funding allowed per player, but would have to go first in the funding round. This meant, if a reptile wanted to force themselves on the boat, they would have to win the auction round, allowing for mammals to react.

Additionally, we replaced some un-interactive mammals’ roles and combined revealing as a good player with more nuanced roles that provided revealing or information at some information or gold cost.

We also added two new reptile roles to make the reptiles in game uncertain.

Playtest 3 (Monday game night – full game)

On Monday, we playtested the game with new printed gold pieces and the auction rule. The table consisted of 2 CAs, 2 group members, and two others. We observed the following reactions:

Players seemed to really enjoy using the gold pieces, commenting on how it looked great visually but also how it felt satisfying to handle.
The auction phase prevented the previously degenerate game pattern of the last funder being able to entirely decide the people who go on the ark.
Players noticed frustrating distributions of roles in a deck with 3 copies of each card. Some configurations completely invalidated our investigation system.
Players commented on how reptiles get found out too easily with investigation. But at the same time, the mammals can’t do anything to oust the reptiles even once they were revealed.
Players disliked that some roles were able to easily reveal and verify their identity without any drawbacks.
Players noted how the rabbit is really strong, and can only ever be counterplayed by the snake.
Players noted that the turtle is necessary so that evil players are not immediately found when showing an evil card.

Iteration – The Companion App, Coup, Limit Whispering

During this playtest, we verified that the mechanics added by the previous iteration worked. The auction mechanic both prevented a previous game-breaking interaction and added another layer of depth that appeals to our primary target audience. Since the purpose of the auction is not immediately obvious, players need to think about the consequences of both winning and losing the auction.

Beyond that, we realized that we still had major system-level issues that we needed to triage before the next playtest.

Firstly, there were a lot of combinations of role cards that would invalidate the investigation mechanic in our game. A reptile with 3 reptile cards cannot effectively bluff at all. A player with the same real and fake identity can essentially prove their identity in one investigation. To address this, we needed a way to enforce specific conditions for what can’t happen. Our solution to this was to make a companion app to distribute the roles. We programmed it to both enforce the no-duplicates rule and draw each players’ fake identities out of a virtual deck without replacement. This way we could adjust the number of reptile and mammal cards in the deck to lower the chance of drawing unfavorable sets of identities.

Then, we needed to solve the issue of there being nothing the mammals can do against a fully revealed reptile. To do this, we thought about the relative value between the money of the reptiles and mammals. In order for mammals to block reptiles from entering the final ark, they need to fund 4 people when reptiles only need to fund 1. That makes it so that, in the endgame, reptile money is worth 4x mammal money. To let the mammals catch up, we needed a mechanic that could remove money from reptiles at a cost that was less than 4x their gold. This led to the coup mechanic, set to remove a target’s gold costing 1.5x their gold. The threshold was set to 1.5x because we thought it would generate a significant difference if used correctly, plus it’s an easy number to calculate mentally.

To address the issue of revealed roles being too powerful, we realized that it would take more than nerfing the cost necessary to perform a reveal (although we did that as well). The system-wide problem was that once a player is revealed, all of the good players can whisper their role to the revealed player. This also completely invalidated the core dynamic that the investigation mechanic is trying to create – that information should be dubious and should need to be corroborated between different players. To explicitly ensure this doesn’t happen, we banned whispering to revealed roles.

The final issue was regarding how the snake’s ability to disable the rabbit’s instant win power was necessary for the reptile team to even have a chance at winning the final ark. The problem was that if a necessary power is tied to a role, we were forced to include that role in every game. With only two reptile slots, the number of possibilities is drastically limited, which reduces the challenge of figuring out which roles are in play. To address this issue, we factored out the power of disabling a mammal’s power to a reptile team power. We reckoned this wouldn’t be too overpowered because it’s not necessarily good to disable the mammals’ powers (many of them are actually drawbacks).

Playtest 4 (Tuesdays during 4A – 60 min, one round) & Playtest 5 (Wednesday game night, full game)

Playtest 4 was conducted with players who self-identified as social deduction players from in class, all of whom had never played the game before. Playtest 5 was conducted with players with varying degrees of social deduction experience at board game night. Two of those players had played the game before. There were many observations that were shared between the playtests:

Players found the vocabulary in the rulebook confusing.
- They took a long time figuring out which token corresponded to which purpose.
- Players without a social deduction background didn’t understand what actions like “reveal” meant.
Players found ambiguity in the rulebook’s wording.
- They were unsure if “players may investigate once per turn” referred to one investigation collectively or one per player.
- The rulebook didn’t explicitly state who should start off with the Auction Token.
Some rules were explicitly stated, but buried too deep within the wall of text for players to find and reference.
- Players bid in free order instead of clockwise like we stated.
- Players suggested house rules like using rock-paper-scissors to break funding ties.
- Players did not realize they could distribute funding to themselves.
Reptile players were able to play with different strategies. Some aggressively demanded claims and gave away their resources to act good. Some stayed quiet and hoarded gold. Some realized that they could reveal themselves once they had accumulated enough gold.

The main difference between the playtests is the composition of players and their experiences at the table, and that created very different dynamics at the table.

In playtest 4, all players shared equal roles in the discussion, so no single player necessarily dominated the discussion. In this playtest, players didn’t raise concerns about the amount of gold available.
In playtest 5, the players with more social deduction experience dominated the discussion and persuaded the inexperienced players to subscribe to their theories. This led to the table collectively waiting for one person to mathematically justify the outcome of the game. By the time they had figured out the math, they realized that mammals lacked money to win from a very early point in the game.

Iteration – Visual clarity in the rules, gold balancing

The main takeaway from the shared feedback was that our rulebook lacked visual clarity. This hurt the playability, which increased the time barrier before people started understanding the game and having fun. To address this, we added the following changes to the rulebook:

We made a small flowchart of the game’s phases.

We made a cheat sheet of the objectives, roles, and actions in the game.
The game flow part of the rulebook was reorganized so different phases had cleaner visual separation.

It was really good that the reptile team played with different strategies because it showed the strategic depth of our game. That’s a dynamic that we wanted to maintain. And in order to do that, we had to make sure the amount of resources is balanced in such a way that there is no clearly dominant strategy. The issue of the mammals lacking gold needed to be addressed because it leads to a solved game state in certain endgame situations. This completely took away fun through both discovery and fun through challenge because the outcome was already decided. Not only that, but a limited pool of gold heavily restricted the actions that mammals can take because they need the money simply to vote. In future iterations, we made starting gold a variable that we could tune.

Playtest 6 & Playtest 7 (Thursday night at Murray, two full games)

Playtest 6 was conducted with one player who played in Playtest 2 and five new players who had all played social deduction games before and were within our target audience.

After Playtest 6, the majority of players expressed strong interest in playing again, so we conducted Playtest 7 right after. One player had to leave, so one team member played in this round. All five other players stayed.

Because mammals had easily won in Playtest 6, we reduced the amount of starting gold from 8 to 6, but made no other changes between playtests.

Our key conclusion from these playtests was that mammals had a strict advantage regardless of the starting amount of gold because there was not a strong disadvantage of freely sharing information. In both playtests, mammals easily won by having the first two arks succeed because with every investigation, all players would choose to share the exact roles they were shown. This forced all players to publicly claim their roles, lest they be determined evil. With that much information, and the ability to double investigate a player, mammals were able to easily identify the reptiles and coup them to prevent them from boarding the boat. The threat of the silence token was ineffective, because the key good role to be silenced only applied to Ark 3, which the game always ended before.

We also observed that players wanted to continue playing our game after playing once and that playability drastically improved upon replay. We did not plan on conducting multiple playtests, but after strong interest from the group, we happily played again. This second playtest was 30 minutes shorter than the first, with only small clarifications needed from referencing the rulebook.

We also observed the following:

Two reptiles both publicly claimed the same mammal role because they couldn’t coordinate beforehand, which led to them easily being found out.
In playtest 6, when players started round 1 with 9 gold, players would investigate and coup indiscriminately. Despite double-investigating their teammates and even successfully couping one of their teammates, the mammals still easily won and had ample gold remaining.

Iteration – Final Assassin Mechanic

To fix the key issue of mammals winning by freely sharing information, we wanted to create a stronger incentive for mammals to want to obscure the information they share. However, we strongly disliked the idea of adding a mechanic that would allow the reptiles to win purely based on luck (eg: having reptiles win if they can guess a single player’s identity would give reptiles a 25% chance of winning from pure luck). We landed on the following: reptiles win at the end if they can correctly guess all of the mammals’ true identities. Thus, mammals would not want to publicly reveal their true identity, while reptiles cannot win off pure luck.

This change also addressed an issue from Playtest 5 where a revealed reptile felt bored because they weren’t able to make more actions. However, with this new mechanic, they would have something to do even after revealing.

In order to make it more difficult for reptiles to guess mammals’ identity, we made the Rabbit’s sole ability to Obscure (choose to hide their real identity during investigation) and moved all of the Rabbit’s abilities (including Obscure) to the Elephant. Allowing an additional mammal to Obscure allows mammals to better hide their identity, while moving the Rabbit’s automatic success power to the Elephant addresses the concern of the Rabbit role being unfairly powerful. The automatic success is hindered by the fact that the Elephant is also heavy.

Additionally we added the following changes:

To prevent reptiles from claiming the same good role: reptiles share their fake and true identities with each other during the night phase.
To avoid excess couping and investigations: players start with 6 gold at the start of the game.
- We chose not to explicitly disallow double-investigations because with the new guessing mechanic, double-investigations to explicitly confirm a role could be disadvantageous to mammals.
To speed up learning the game, we created a Quick Start / Cheatsheet guide.

Playtest 8 – Final (Friday at 11:59 PM in Murray)

Timestamps are noted below in the timestamp section, and also in the YouTube description.

Our final playtest occurred at midnight on Saturday, with 3 new players who were part of the target audience, 2 players who had played in Playtests 6 and 7, and one team member. The new players led the initial set up and clarified questions by consulting the rules.

The final playtest went extremely well. Players were able to get up setup within 10 minutes and questions about the rules were answered through referencing the rulebook. Additionally, players had more fun through challenge and fellowship as the game felt more balanced, narrative through the dynamic element of the animals theme, and discovery with the additional mechanics and higher difficulty.

There was a clear shift in player dynamics due to the new mechanic of reptiles winning through guessing mammals’ identity. During the rounds, players, including mammals, now rarely shared full information, making it difficult to determine the reptiles. This added to the challenge of the game because mammals did not have an easy win strategy. It also added to discovery because both reptiles and mammals had more potential strategies to explore, rather than a single key strategy becoming paramount. Finally, throughout, players enjoyed immersing themselves in the theme of the game, often referring to each other by the role name, even enacting certain roles. Players also expressed joy with the high fidelity of the individual game pieces.

After the game, the players who had played Playtest 6 and 7 said the game felt much more balanced comparatively and even suggested playing again. The suggestions we received were around further improving role agency and rules clarifications, discussed below in more detail. They spent more time thinking about strategy in this playtest, leading to a much more satisfying outcome. Playtests 6 and 7 were the most expressive, but it was because the players were oversharing. Playtest 8 is a more balanced and strategic game.

Future Changes

For our final playtest, parts of the rulebook were still unclear. This includes the bidding mechanic in the auction phase, parts of the preamble that should be reordered, and when a certain role receives money, which have been clarified in our final rulebook. They also suggest making the role of the companion app be more specific, which is also clarified in the final rulebook.

If we had more time, we would want to also test the following role balance changes. The player who played the fox expressed that they didn’t feel like they had as much choice in the game regarding their role. To fix this, we would like to test a role where the fox is able to fully replace their identity with a reptile card, allowing them to fully hide their identity. We also want to change the chameleon role, as in the current iteration, it is significantly weaker than the other two reptiles.

Final Formal Elements

Players:
The game is for 6 players, four as Mammals and two as Reptiles.

Procedures:
Starting actions include game setup and reptiles learning who each other are. The core loop proceeds as follows:

Discussion phase — players investigate, discuss, and coup.
Nomination phase — players auction the amount of gold they want to spend.
Funding phase — players allocate gold to those they want on the ark.
Expedition phase — the selected players each submit either a sail, heavy, or sink token in a bag.
Night phase.

The loop continues until two arks have either sailed or sunk, or three rounds have passed (whichever comes first).
The Reptiles also have a resolving action: guessing the roles of each of the Mammals.

Resources:
Each player starts the game with 6 gold, which is used to pay for investigations and to cast votes.
Every player receives 1 gold at the beginning of each round.

Conflict:
Mammals don’t know anyone’s alignment. Players must manage resources to gain information and voting power.
Reptiles are outnumbered.

Outcome:
Mammals win if two arks successfully sail (or if one ark sails and the elephant is on the final boat).
Reptiles win if two arks sink. The outcome is zero-sum between the teams.

Final Components

Instructional Video

Note: in the physical copy of our game, the cards / player mats / board were printed on cardstock, the markers and gold were 3D printed, the tokens were laser cut (with stickers added on top for investigation / auction / silence), and the passports were ordered.

Print & Play

Print and Play link.

The Print & Play consists of the following materials (in order): marketing materials (box top and bottom), tokens, play mats, main board, passports, identity cards, rulebook, and cheat sheet.

Playtest Timestamps

Final Playtest:

48:23 – “Finding the cold blooded killers” – Players made puns with the theming, showing that they enjoy the aesthetic.

1:36:04 – Emergent strategic depth: a player was forced to investigate the revealed cat to waste their money.

1:33:28 (1 min) – Hippo card and a crocodile card. “This is the hardest version of the game I’ve ever played” “What if he has the turtle” – callback to when 3 ppl had the turtle

1:55:52 – Elephant puts in a SINK to hide that they are HEAVY. People think he is trolling but hides his identity as the elephant so he could use his instant win tool power later. This is also talked about in the post game interview.

Post-Game Interview (2:00:20) : At the end of the game, reptiles had thought they had won, but the new elephant mechanic game came into play allowing mammals to win. Jay’s commentary showed the new elephant mechanic was effective as because the role was heavy, it would have revealed him as the elephant, except he decided to sink both boats instead. He only needed to do this because he needed to hide from the reptiles, which he successfully did.