P2: “The Sixth Face” (Team Bathyergus)

Team Bathyergus: Xingyi, Ellie, Claire, Dana, Hazel

Artist’s Statement

The Sixth Face is a single-player puzzle videogame for the Playdate system centering on Vincent van Gogh’s experiences during his period at the Saint-Rémy asylum. The player explores a mechanical cube, with each face a different puzzle built from a painting he made while institutionalized. Designed for young players and any puzzle enthusiast, the game can be experienced both independently or with adult guidance, encouraging an appreciation for art while exploring the artist’s stylistic and creative choices.

Gameplay is designed for the player to interact with the artwork while exploring the narrative. Each level converts a different Van Gogh painting and its emotional register into an interactive puzzle. Starry Night converts the view out of Van Gogh’s bedroom window into a sliding-tile puzzle made up of the window panes. A Corridor in the Asylum loops a maze endlessly, disorienting the player. Almond Blossom asks the player to manipulate broken branches back onto a trunk. Haystack under a Rainy Sky prompts the player to rediscover the painting’s unusual  composition and perspective. Self- Portrait with Bandaged Ear allows players to immerse themselves into an artistic world both visually and auditory in order to better understand Van Gogh’s experiences. The interactive, art-based mechanics of each level create a dynamic of deep interaction with Van Gogh’s inner thoughts and an aesthetic of Discovery, Challenge, and Narrative. 

Logistically, the game assumes no prior knowledge of Van Gogh and keeps each level self-contained. Rather than presenting a biography, it aims to evoke an impression of Van Gogh’s inner mental experiences. This approach is further enforced by Playdate’s one-bit screen limitation, and echoes Van Gogh’s personal artistic ideals: impression, not recreation. The experience is whimsical and reflective, open to players producing their own interpretations. We hope that each player closes the box feeling both challenged and introspective, having spent so much time in the digital asylum space exploring the mind of the artist.

Model of Chosen System

Initial Decisions About Formal Elements and Values of Your Game

Initially, we wanted a game about looking inward—a dive into the mind rather than across an expansive map, since fiddling with a puzzle is similar to the process of understanding a person. This set our primary types of fun—Challenge from puzzle-solving, Discovery as the priority, and Narrative as the payoff—alongside three values: mechanics as metaphors (each puzzle should capture the feelings associated with its source painting rather than just depicting it), impression over recreation (like Van Gogh’s work, whimsical and unafraid of disorientation), and openness to interpretation (some evocative narrative for anyone slightly familiar with Van Gogh). Our biggest worries were structural: with five designers each building a level around a different painting, the puzzles need to be chronologically coherent, stylistically cohesive, sufficiently varied in puzzle type, scaled along the same difficulty curve, etc. We therefore committed to the puzzle cube as the connecting thread—each solved face draws the player physically closer to a completed cube, and to the impression that you’ve closer to the core of a person’s mental state. To accomplish this, we came up with an idea containing the following elements:

• Objective: Solve the mechanical puzzle box—each face delves deeper into Van Gogh’s mind.
• Players: Single-player. Targeting children, casual gamers, or puzzle enthusiasts; anyone even slightly familiar with Van Gogh.
• Boundaries: The mindspace of Vincent van Gogh during his stay at Saint-Remy (each level a painted scene formulated as a puzzle, bounded by the faces of the box rather than geographic space).
• Resources: The PlayDate’s full control set (d-pad for navigation, A/B for interaction, crank for rotation/manipulation of scenes), plus in-level clues (e.g. image preview of Starry Night, musical/visual hints, etc.).
• Procedures: Each level draws its procedure from its painting: maze requires navigation and search, tile puzzle requires sliding and image recreation, scene-matching puzzle requires viewpoint alignment, music puzzle requires pattern recognition, etc..
• Conflict: The player versus the box—struggle to solve and understand the box (and therefore the person).
• Outcome: Each solved face yields a new deeper understanding of a painting—the game’s finished state also doubles as Van Gogh’s creative survival.

Space and Architecture

Our game treats space not simply as a setting, but as a way of representing Van Gogh’s memories, perceptions, and inner mentality. The cube functions as a map of this mental space: although initially appearing small, each face opens into a much larger and more vivid experience. In this sense, the gamespace becomes a metaphor for the mind. Playing the game allows players to encounter fragments of Van Gogh’s life through movement, observation, sound, and interaction rather than through exposition alone.

The Corridors and Haystack puzzles reconstruct aspects of Van Gogh’s physical environment. In the corridor puzzle, players navigate the repeating architecture of Saint-Paul-de-Mausole without access to a map. Corridors connect back to themselves, doors appear similar, and some spaces cannot be entered. This creates feelings of confinement, uncertainty, and disorientation. We initially designed the maze to be more difficult because we wanted players to experience confusion and frustration—what Van Gogh might feel when experiencing psychosis. However, we had to balance this emotional intention with playability: if players become completely lost, they would disengage from the experience rather than reflect on it. The level therefore uses subtle environmental information—such as the fountain, patients, nurse, fire, and landscape—to help players gradually form an incomplete mental map. The Haystack puzzle presents a different form of spatial limitation. Instead of a confined interior, the player wanders a field near the institution. At first, the static landscape appears bland and unremarkable, dominated by a haystack in the center of the field and dotted with dreary puddles. The player walks around the field, using a sketch to interpret the landscape and discovering reflection and perspective through the silhouettes of crows in the puddles, eventually finding an interesting point of view to paint. There was little happening around the asylum, which was in the countryside, and both puzzles reconstruct the restricted environment surrounding Van Gogh during this period: the corridor produces confinement through architecture, while the field creates isolation through open space and apparent monotony.

The Starry Night and Almond Blossom puzzles operate differently. Their spaces do not primarily reconstruct a literal location; instead, they take place within the visual worlds of the paintings. Players rotate tiles to reconnect the movement of light in Starry Night and rotate branches to join blossoms back to the tree in Almond Blossom. These interactions make the paintings feel alive. The flowing light, turning branches, and emerging flowers allow players to participate in the paintings’ movement rather than viewing them as static images. These levels represent Van Gogh’s imaginative and artistic inner world: ordinary skies, trees, and landscapes are transformed through rhythm, motion, and expressive brushwork.

The Self-Portrait with Bandaged Ear puzzle creates a third kind of space: a musical and textual play space. Each piano note represents its letter name. Players listen to sequences of notes and sound effects and reconstruct words associated with its aftermath, such as “bandage,” “ear,” “face”. This mechanic adds variety to the game by shifting the player’s attention from visual navigation and object rotation to listening, interpretation, and language.  This puzzle also differs in perspective. In the other four levels, players experience each environment through an implied first-person viewpoint: they see what Van Gogh might have seen, but they do not see his body. In this puzzle, however, players directly manipulate a visible human figure. Van Gogh’s body becomes part of the play space because the painting is a self-portrait. This shift from seeing through his perspective to looking at his represented body reflects the painting’s focus on how he saw and presented himself. Injuring his ear contributed to his decision to enter Saint-Paul-de-Mausole, so this also serves as a prologue.

The five faces can be explored in any order. Each level therefore feels like an independent fragment rather than a chapter in a fixed chronology. This resembles the nonlinear structure of memory: different places, sensations, and experiences exist beside one another without forming a perfectly ordered narrative. Small environmental connections make the fragments feel as though they belong to the same mental landscape. For example, the corridor offers glimpses of the field, while the field allows players to see the institution from outside. Corridors and the Haystack share the same suspense background music, while the Starry Night and Almond Blossom use the same more artistic music. These relationships remain intentionally subtle. (This unordered level design is also discussed more in later sections.)

Ultimately, the cube (and also the physical playdate ) is small enough to hold in the player’s hands, but every face unfolds into a larger experiential space. Its architecture suggests that a mind can contain disconnected yet intensely vivid memories. By entering these spaces, players do not literally experience Van Gogh’s mental state, nor do they “solve” his struggles. Instead, they encounter interpretations of the environments, sensations, and artistic processes that shaped this period of his life. The game asks players to see what he saw, move through spaces inspired by those he inhabited, and interact with the visual rhythms through which he transformed experience into art.

Design Sketches/Mockups

Scope of the Game

Early in development, we decided our P2 submission should be an MVP of the complete Sixth Face rather than a very-small-portion polished slice. The game is a box with a fixed number of faces, and the central promise that each level translates a different painting and puzzle type that draws closer to the box’s core can’t be demonstrated by one face alone, no matter how refined. So we built five levels around five paintings — Starry Night, A Corridor in the Asylum, Almond Blossom, Self-Portrait with Bandaged Ear, and Haystack under a Rainy Sky — each taken through its own playtest-and-revision loop, then combining into a cohesive psychological/narrative game.

Testing & Iteration History

Guiding Questions

1. How long does the game and each level take each player?
2. How do you feel about the (1) difficulty and (2) thematic/narrative flow of the puzzles?
3. How much familiarity do you have with Van Gogh or any of his works?
4. What are the aspects of the game that you find most challenging?
5. If you were the designer, what would you keep and/or change? 
6. How familiar are you with the PlayDate control system?
7. (Level-Specific) Do you have any fundamental knowledge about music notes?

Independent Per-Puzzle Playtesting

Each puzzle, before integrated, had at least three rounds of playtesting.

Almond Blossom

	Findings	Improvements + Decisions
Playtest 1: CS247G Peer	Some wording was unclear, especially references to “blooming.”	Adjusted puzzle opening to make goal clear.
Playtest 2: Non-CS247G Friend	The circuit-like visual is appealing to someone who is familiar with similar systems.	Kept circuit-like visual style.
Playtest 3: Butch	Every piece seems to have only one logical connection—the game might be too easy. Corner pieces make subsequent placements obvious.	Chose pieces/orientations wisely and didn’t overuse corner pieces.

Haystacks under a Rainy Sky

	Findings	Improvements + Decisions
Playtest 1: CS247G Peer	Did not realise the role of the crank in adjusting the view.	Added text explaining the usage of the crank and changed the sketch to force players to experiment with the crank.
Playtest 2: CS247G Peer	Did not understand that the player needed to match the sketch to the real-world view.	Added cutscene with poem to level to explain backstory and assign the task of finding the view shown in the sketch.
Playtest 3: Butch	There’s noticeable input and rendering lag on the PlayDate.	Changed from rasterizing the entire scene to calculating only the positions of key points.
	The 3D open-world field area is too large, leading to slow rotation speeds and tedious wandering.	Reduced field size.
Playtest 4: CS247G Peer	Did not recognize that the distance measurement shown in the hint showed how far the player was from the target location.	Adjusted the wording to make it clear that the hint shows the distance to the target location.

Self-Portrait with Bandaged Ear

	Findings	Improvements + Decisions
Playtest 1: CS247G Peer	Accessibility: Tone-deaf individuals, those with auditory disabilities, or even just those without musical knowledge may face difficulties when playing the game due to the high reliance on music notes.	Added musical and sentence hints to the game (more accessible without music knowledge). Added sounds to the letters so that players are able to match the music.
Playtest 2: CS247G Peer	Wandering around the room without knowing specifically where to go caused the players to be confused about where the puzzle starts.	Added arrows that guide the players to the canvas in order for them to enter the puzzle.
Playtest 3: Butch	Play the corresponding note immediately when a player selects/presses a letter key. Allow players to listen to what they have typed compared to the original sample before submitting.	Each letter/note pressed played the corresponding sound so that players are able to match their notes with the given hints.
	In the “bandage” puzzle, keep both “A” notes within the same octave, as detecting octave leaps on 8-bit audio is too difficult for average players.	Adjusted notes to be the same octave, allowing players to easily match the notes with the hint.

Corridor in the Asylum

	Findings	Improvements + Decisions
Playtest 1: CS247G Peer	The map is too complicated.	Made the map simpler so it feels tractable.
Playtest 2: CS247G Peer	Players don’t know what to do after encountering a location (don’t know what the goal of the game is).	Added a background scene to know the story behind the game. Explained the goal, to navigate to the infirmary. After the level finishes, showed the map of the game.
	Windows and doors of the same design make it really hard to navigate. Unclear about the floor and the walls.	Added a list for the Map symbol. Added more variations for different doors and windows.
Playtest 3: CS247G Peer	Players finished too quickly without fully exploring the map.	Added a system for forced exploration where they have to visit three scenes.
Playtest 4: Butch	Provide a narrative justification for displaying map tracking/stats at the end.	Added more narrative explanation.

Starry Night

	Findings	Improvements + Decisions
Playtest 1: CS247G Peer	Player felt the direction of the arrows was unintuitive/should be flipped.	Changed slide mechanic to be in reversed direction (as if moving the blank space) and changed blank space representation to be more clear.
Playtest 2: CS247G Peer	Player found it hard to imagine the final image, especially in black-and-white.	Added a reference image, held behind a button-press.
	Complexity of 4×4 grid was ideal and looked more like window panes.	Abandoned consideration of a mandatory 3×3 grid.
Playtest 3: Non-CS247G Friend	New sliding direction is unintuitive (contradicting first playtester’s opinion).	Option to reverse controls direction, but only in “Help” menu so as not to confuse.
	Narrative payoff doesn’t really make sense.	Change it so the window panes are disorienting and jumbled, but the crank leads to a better shooting star animation, which lights up the landscape of the puzzle that was just solved.
Playtest 4: Butch	Puzzle is quite difficult without support for some players.	Followed many of George Fan’s Plants vs. Zombies onboarding techniques (See Onboarding section.): Gave information about crank only later in game when salient, added an only-if-desired hint option, and added an only-if-confirmed “help” option that simplifies the puzzle to a 3×3 by solving the outer edge.
	Crank/light mechanic does not make sense and requires onboarding.	As mentioned above, moved crank prompt to end of puzzle as a reward for solving.
Playtest 5: Butch (August 13)	Wasn’t clear that the A button showed the reference photo.	Rephrased A to be “Help,” which encompasses reference photo, hints, help and controls changes.

Onboarding Considerations

With five levels using five entirely different puzzle types and control schemes, we couldn’t front-load a single tutorial; each face had to teach itself. We therefore relied on short prompts (never too much text, as is shown to be most effective) at level entry and on unobtrusive, passive messaging that doesn’t break flow. However, we do try to familiarize users with the Playdate control that is shared in all puzzles. Before seeing the main cube, they have to use the crank to scroll down an opening narrative, and press A to continue. In the main cube, they will need to use the D-pad to rotate the cube and crank again to enter the level. 

Playtesting showed where this wasn’t enough. For example in Starry Night, two pain points were solved via onboarding changes:

• Timing of Information:  Some players were confused by a crank mechanic being mentioned at the start of the puzzle but then the crank not being used until the very end. As Geroge Fan mentions in his GDC session, information should be given when it is most salient. So we changed the ordering such that the initial screen only explains the tile puzzle, and only upon solving does the game prompt the player to turn the crank to reveal a beautiful Starry Night shooting star visual.
• Adjustable Level of Difficulty to Learn Strategy:  Our second lesson concerned hints. Playtesting established that Starry Night needs a strategy hint — the 15-puzzle technique of solving the ring opposite the final empty corner, reducing 4×4 to 3×3 to 2×2 — since without it, completion times ranged greatly depending on familiarity. 
• Adaptive Messaging:  Unobtrusive and Only-When-Desired Hints:  But the same playtest revealed the opposite problem: our playtester twice opened hints he hadn’t meant to see (“Oh, I did not want to look at that, actually”), spoiling content he wanted to earn. Hint activation was too easy to trigger accidentally, and players couldn’t tell what kind of help they were about to receive. We rebuilt Starry Night‘s hint system around adaptive messaging — only show messages if players need them, but leave room for exploration. The hint menu now offers two distinct options: a strategy hint that teaches the solving technique, and a “Help Me Solve” option that makes the puzzle easier directly. Each states what it will do before it does it, and requires confirmation. A player who wants to learn the trick can take it; a player who wants to preserve the challenge can back out; and neither can stumble into a spoiler by pressing a button they didn’t mean to press. This player-choice strategy is present in many levels: In each of the puzzles, players can choose to use the hint. In the Corridor level, they can press B for more explanation, in the Ear level, they can have textual hints in addition to the pure musical ones, and in the Haystack level, players can use the hint to see the distance between to their destination.

Demonstration of Experience

Notable Moments:

• Timestamp 01:38:  In the Corridors puzzle, the player sees the pop-up “Press A to Enter” the first time they encounter a door, and they successfully enter the door, and then from that point on the player enters doors without any help. This follows George Fan’s Plants vs. Zombies onboarding wisdom that “You just have to get the player to do it once, and then they know how.” We also use adaptive messaging, and after the first instructional pop-up is successful, it then changes to just a small “A – Look” in the corner.
• Timestamp  04:45:  The player watches narrative scenes at the start of the Haystacks puzzle that give context on why Van Gogh is now wandering into a haystack scene and solving the puzzle. This follows the thematic coherence heuristic, where mechanics, visuals, and story reinforce a unified theme. It also follows George Fan’s wisdom to use minimal text and prioritize visual, already-known-by-player information.
• Timestamp 20:14: The player completes the Starry Night puzzle and is clearly proud, stating “Boom!” Only then does the game give crank directions, stating “Congratulations! Now turn the crank to see a Starry Night!” This is a “classic game puzzle”, and this joy is a clear sign of Marc LeBlanc’s “Challenge” type of fun at work—players feel a sense of pride and accomplishment upon finishing the hardest level of a game—as well as George Fan’s wisdom to only deliver information when most salient (not mentioning the crank at the start of the game, when it would confuse players).
• Timestamp 27:20: The player reads the clue out loud, then uses it to solve the hardest level of the Ear puzzle. This is a great example of designing and adjusting good puzzles—because not all players might be able to solve using only extensive music knowledge, we “leave breadcrumbs” in other formats/domains like verbal hints and visual musical staffs, changing the amount of directness to hopefully eventually lead the player to the solution. The player even then says “Ohhhhh, I know this!”—Literally the ideal “V-8 response” described by Bob Bates in his GDC writing.
• Timestamp 29:10: The player immediately latches onto how the Almond Blossom puzzle works because the un-connected branches visually make it clear that they are supposed to be connected. It uses the graphic design concept of “Closure” to immediately realize what the branches should look like connected. The player also states out loud that “I like these kind of games”—Marc LeBlanc’s “Challenge” type of fun at it again.

Uploaded Build of Game

Itch.io Link:  https://xingyi-zhang.itch.io/the-sixth-face

Gameplay Video/Trailer

Extra Feedback & Testimonials from Playtesters

• Already have interested players both collecting and downloading it on itch.io within one day of publishing! (See above.)
• “I liked your game!”
• “I like the music they used.”
• “Ooh these games are good!”
• “Oh I like these kinds of games!”
• “It’s a nice game to just spend lots of time with.”
• “Very technically impressive.”
• “I had a lot of fun with the music puzzle.”
• “It’s a fun short puzzle game.”
• “Starry Night and Almond Blossoms were my favorite.”
• “This is fun! This is my first time being able to complete [one of these kinds of games]!”
• “It can be sold!”

Behind-The-Scenes Notes:  Difficulty Curve

To gauge puzzle difficulty, we decided to gather per-puzzle timing statistics.

Puzzle	CS247G Peer (8/11)	Butch (8/10)	Non-CS247G Friend (8/11)	Non-CS247G Friend (8/11)
Haystacks	2:50	5:36	4:35	2:50
Blossom	5:02	1:50	2:54	3:55
Echoes of the Canvas	2:56	6:15	4:30	3:10
Corridor	1:12	2:27	6:26	8:00
Starry Night	30:59	8:33	15:00	7:32
Total	42:59	24:41	33:25	25:27

Analysis shows the puzzles at reasonable similar durations, with the median Starry Night time being slightly longer, but the outliers do not cause alarm, as hints and “Help Me Solve” features were added to mitigate this and bring the median closer.

Behind-The-Scenes Notes:  Order of the Game

We considered three ways of ordering the five puzzles: chronology, difficulty, and spatial or narrative connection.

A chronological structure would arrange the paintings according to when they were created or when the related events occurred. However, this conflicted with our concept of the cube as Van Gogh’s mental space. Memories are not necessarily recalled chronologically; they can remain simultaneously accessible, fragmented, and disorganized.

We also considered ordering the puzzles by difficulty. Our playtest results showed a general difference in completion times: Starry Night was significantly longer than the other puzzles. However, difficulty was highly subjective. For example, completion time for the corridor ranged from 1:12 to 8:00, while Starry Night ranged from 7:32 to 30:59. During playtesting, we observed that players responded very differently depending on the type of puzzle. A player who easily understands spatial navigation may struggle with sliding tiles, while another may have the opposite experience. Therefore, a fixed difficulty progression would not accurately reflect every player’s experience.

A third possibility was to connect the puzzles through space or narrative—for example, moving from the field in Haystacks to the institution in Corridor in the Asylum, possibly using cutscenes between them. Although subtle spatial connections remain in the game, enforcing this sequence would create an artificial dependency between puzzles – which all have completely different mechanisms and the narrative doesn’t not depend on each 

We therefore decided not to restrict the order of exploration. Players rotate the cube, select any available painting, and move freely between its five faces. This nonlinear structure supports the cube’s representation of mental space: separate memories and perceptions remain accessible at the same time, even when their relationship is incomplete or disordered. It also gives players agency to follow their curiosity rather than a prescribed sequence.

Players can leave a puzzle and return to the cube through the Menu at any time. This prevents one difficult puzzle from becoming a barrier to the rest of the experience. Players may explore another face and return later, allowing challenge to influence their individual route without determining a universal order. The structure therefore uses the cube as an open hub that balances thematic disorder, player freedom, and differences in puzzle-solving ability.

If This Were a Full Release…

If this were a full release, we would consider extending the game to encompass even more of Van Gogh’s works–Van Gogh painted ~150 works at Saint-Rémy, so this setup could absolutely scale. This could mean more puzzles per face out of the six, or something more creative like additional faces as you shed the first layer of the box. The game could also be divided into chapters with associated themes or timelines–We would focus on his time in the asylum, his move to Paris afterwards, etc. More puzzles would also mean more cross-puzzle mechanics, where choices in one face/level affect scenarios in later ones. Finally, a color-platform or non-one-bit release alongside the PlayDate original would be exciting, as it could be more colorful like his artworks, incorporate color into game mechanics, and be more accessible to non-PlayDate users.

Accessibility Notes

The Playdate’s one-bit screen shaped our accessibility choices from the start. Without color to distinguish walls, corridors, paint colors, etc. (something the maze level playtesters struggled with), we had to make changes in each level to accommodate the fact that color wasn’t there to help: clearer symbol differentiation and an in-game map in the maze level, texture rather than color and a reference image in the slide-puzzle level, etc.–which incidentally makes the game fully colorblind-accessible.

Our other biggest lesson came from the Bandaged Ear level, which asks players to identify notes by ear. Playtesting showed the puzzle gated on musical training rather than reasoning, so we added a note key and more accessible hints. It would remain inaccessible to deaf and hard-of-hearing players, so a full release would need a visual alternative.

Art & Music Credits

• Corridor in the Asylum (September 1889), The Metropolitan Museum of Art. View source.
• Self-Portrait with Bandaged Ear (January 1889), Courtauld Gallery, London. View source.
• Almond Blossom (February 1890), Van Gogh Museum, Amsterdam. View source.
• The Starry Night (June 1889), The Museum of Modern Art, New York. View source.
• Wheat Stack under a Cloudy Sky (October 1889), referred to in the game as Haystacks under a Rainy Sky, Kröller-Müller Museum. View source.

• Gnossienne No. 1 by Erik Satie. Performed by La Pianista and licensed under CC BY-SA 3.0. Wikimedia Commons source.
• Clair de lune from Suite bergamasque by Claude Debussy. Public-domain MIDI transcription maintained by Keith O’Hara, Mutopia Project, reference 1778. Rendered in-game using Playdate synthesizers. Mutopia Project source.

AI Acknowledgement

AI helpers were used to develop code and images for the game.