DM3107 - Major Research Project Part 2
Word Count: 2200
21/05/2026
Introduction
Galatea is a story-driven top-down action-adventure game developed to explore how dungeon design can balance challenges and engagement within a controlled gameplay environment. Inspired by the structure of dungeons found in The Legend of Zelda: A Link Between Worlds (2013), this project investigates how environmental design, pacing, and mechanical progression can be used to create gameplay experiences that remain challenging without becoming frustrating. The project focuses specifically on examining how players interact with an environment, encouraging exploration and teaching mechanics through interaction and repetition rather than direct instruction.
The initial stage of the project involved research into level design theory and structural principles commonly used within Zelda dungeon design, exploring the balance between player engagement and challenge. Attention was given to the concept of dungeons as carefully paced learning environments, where the level naturally guides the player, and mechanics are introduced in a controlled manner before they gradually increase complexity. Because of this, there is an intentional lack of direct guidance. These findings became central to the design philosophy behind Galatea and informed both the structure of the dungeon and the implementation of gameplay systems.
As a game that incorporates various puzzles into levels, this project is aligned with the Sustainable Development Goal 4: Quality Education (2012). Puzzle games, such as Portal 2 (2011), offer a learning opportunity through boosting creativity and flexibility in students (Moffat et al., 2017), making them a possible tool to put students in a state of mind for learning.
The outcome of this project was a polished vertical slice that was displayed at the University of Winchester’s digital design exhibition, Code & Canvas. This demo would expand the design principles explored in the initial prototype with a stronger focus on enhancing the overall user experience.
Artificial Intelligence Disclaimer
This project benefited from the use of generative AI. Due to an industry focus on 3D modelling and game design, and this project being a solo endeavour, there was a distinct lack of knowledge surrounding C# that would render some of the more advanced mechanics to be too far out of scope to include in the project. This does not harm the project from a game design perspective as it mirrors industry standards of a designer providing specifications to a developer. It is important to note how the AI was barred from making decisions on the creative level.
It would have been preferable to work alongside a human developer. However, due to peers having obligations to their own solo projects, this had to be avoided. It is also important to consider how the use of AI goes against SDG 4 (2012) due to how it inhibited the overall understanding of the code and the learning opportunity programming for oneself would offer. Additionally, it is crucial to recognise the environmental impact AI has on the environment, meaning its use should be targeted and minimal.
Engagement & Challenge
Level Design
Research established that effective dungeon layouts often revolve around a central hub area with branching paths (Larochelle, 2014) that gradually open as the player incrementally overcomes challenges. This structure encourages exploration while maintaining a clear sense of progression by presenting the player with a goal. This diverges from the design of the prototype which followed a linear path from start to end.
Sketches of the Dungeon Layout showcasing a core 'hub'
The core level structure was built entirely in Blender. The method used (Markom3D, 2019) was to use multiple planes connected through a linked duplication (Blender, 2026a) This technique would ensure that each floor tile would share a UV map allowing for tileable textures. The planes could then be shaped into the level using a series of arrays (Blender, 2026).
The level's water is a shader made by AureDevGames.
A primary cause of frustration in the initial prototype was the 3x3 Lights out puzzle that monopolised most of the player’s time. The final version compensated this by lowering the number of switches to a 2x2 grid. Additionally, the initial state of all the switches was now in the off state, a recommendation made during testing to allow the player to better sense how the puzzle operates without direct explanation.
Level Blockout
Old Lights out Puzzle
New Lights out Puzzle
A common strategy used within Zelda Dungeon design is to structure a level around a key item or mechanic. This item both unlocks new pathways and recontextualises previously explored areas. As the player learns the design language associated with the item, they begin recognising situations where its use is required. This principle became central to the design of this vertical slice through the implementation of the Lasso tool and swinging mechanic.
Player Input
The swinging mechanic is a method of traversal that allows the player to cross small gaps and chain swings together to overcome larger spaces. This mechanic enforces a focus on environmental interaction and movement-based exploration. The implementation of this mechanic shows the technical limitations of solo development. Most Unity tutorials surrounding swinging mechanics focus on freeform movement where the players control the angle and momentum directly. However, Galatea required a more controlled and readable traversal suited to an isometric perspective. Implementing this feature required substantial experimentation through Artificial Intelligence.
Swinging Demonstration
Another recurring issue during testing involved player movement. Early versions of the game relied on Rigidbody-based Movement (Unity Docs, 2026b) as it would allow for smoother knockback interactions and more swinging more accurate to physics. However, this caused issues regarding navigating stairs as well as causing the players to get caught on walls, limiting movement and negatively affecting overall player control.
To address this issue, the movement system was rebuilt using a Character Controller (Unity Docs, 2026a). This is another situation where, due to time restraints, AI assistance was utilised to accelerate the process. This resulted in a moment system that felt smoother and more responsive.
Camera
An early design choice was the isometric camera that can orbit the player, found in games such as Breath of Fire IV (2000), this was in response to feedback given regarding interpreting the range of melee attacks. To preserve the isometric feel, the camera turns in increments of 45 degrees, limiting the camera angle to eight cardinal directions.
Due to certain platforming mechanics, a true orthographic camera (Unity Docs, 2026d) would make it difficult for the player to determine distance. Two mechanics would need to be implemented first: a drop-shadow that grows and shrinks depending on the distance from the ground, second: fake an orthographic camera. Faking an orthographic camera can be done through the combination of a faraway camera and a low FOV. Found on Instagram (@thomraktar, 2026), this method creates a small amount of perspective that allows the player to better determine depth.
The isometric camera angle introduced visibility challenges. Initially, walls lowered dynamically when the player entered a room. Although visually interesting, this system became increasingly complex due to the addition of the rotating camera. Instead, a solution involving backface culling (Unity Docs, 2026) was utilised. By orientating wall faces inwards, foreground walls became invisible from the camera’s perspective without requiring additional scripting. This significantly improved performance while maintaining visibility.
Camera rotation and Backface culling
In addition, to aid in keeping the current room be the player’s primary focus, rooms outside the player’s location are hidden from rendering. However, this created harsh visual transitions. To reduce eye strain and minimize potential risks associated with photosensitivity (NHS, 2025), fade transitions were added between room changes. This solution preserved the technical benefit of backface culling while improving accessibility.
Room Transition
Methodology
The methodology used during semester one involved producing a standalone prototype that was tested with a small group of players. The purpose of this testing was to gather insights into design principles found in dungeons from The Legend of Zelda: A Link Between Worlds (2013). This project expanded upon the prototype by directly applying feedback gathered during that early testing.
Development followed an iterative workflow where testing, feedback, and redesign occurred continuously throughout production. Two primary versions of the game were tested. Version 0.1 focused on evaluating base mechanics, while Version 0.2 was a build designed around user experience and visual design to be shown at Code & Canvas.
Data collection combines both qualitative and quantitative methods. Players were recorded navigating the dungeon and subsequently provided feedback. This allowed for determining key strengths and weaknesses within the game design.
In anticipation for the exhibition, the scope of the dungeon was significantly reduced. The final level was approximately half the size originally intended. This decision was informed by the nature of such an event being an environment where visitors would divide their time between multiple projects. A shorter dungeon also helped prevent scope creep and allowed development time to focus on polish and usability.
Previous testing used the Bartle Test (Barr, n.d) to categories player behaviour. While this provided useful insight into player motivations, the Bartle Taxonomy is primarily focused on multiplayer experiences and often relies on restrictive binary questioning. To address this, the project shifted towards using the Quantic Foundry's Gamer Motivation Model (2015). This model measures 12 specific motivations and sorts users into 9 distinct archetypes, allowing for a greater resolution of data compared to Bartle’s four player types (Bartle, 1996).
The four-month development schedule focused on balancing gameplay systems, level production, and iterative testing:
The production schedule accounted for unseen technical problems. Issues such as conflicts between the UI and Player Input system (Unity Docs, 2026c), and the shift between Rigidbody to Character Controller would not have been possible to predict. The additional testing and debugging time were allocated during April proved essential in accommodating these complications.
Testing
Participants involved in testing mirrored a similar demographic to the preliminary prototype testing. All participants were university students ranging from their late teens to early twenties. Testing focused on identifying how players interacted with the environment under minimal guidance while measuring enjoyment and frustration.
Background Questions
Background information of the testers was collected to gauge their familiarity with top-down adventure games and their gamer personality types under the Quantic Foundry model (Quantic Foundry, 2015). Of the six play testers, there were three Bounty Hunters, one Bounty Hunter / Bard, one Slayer / Bard, and one Architect / Skirmisher. Overall being well suited for testing for a demographic interested in story focused, action-adventure with player progression (Quantic Foundry, 2015).
Playtest
Due to the iterative approach to testing, the findings are split across multiple versions: Version 0.1 involving the first two participants in an early build of the project, and Version 0.2 involving the other four participants in preparation of the Exhibition.
Testing revealed several recurring patterns. One of the most consistent observations in player behaviour was the limited use of the rotating camera system. Most players selected a preferred angle and rarely adjusted it during gameplay. As a result, some players missed points of interest or environmental information that was blocked by their chosen perspective.
Additionally, due to the lack of information boxes when collecting an item, many players would obtain their first key and try to use it on the boss door. This issue was intensified if the player had missed the door in a previous loop. This highlights how item descriptions are necessary in providing context to the player, directing them in a way that increases engagement and reduces unfair challenge.
The leading cause of player damage was the double-swing traversal sequence used to obtain the boss key. The automatic latching system occasionally caused players to unwittingly drop into the water hazard. Additionally, players would often align the camera in the direction of motion, exacerbating the difficulty of determining distance compared to a side profile. Going forward, the swinging system should be adapted to be more forgiving because, as posited in The Design of Everyday Things (Norman, 2013), human error should always be considered the fault of bad design.
The redesigned lights out puzzle received a significantly more positive response compared to the original version and contributed heavily to the reduced playtime between versions. However, an argument could be made that the puzzle is too easy to ‘brute force,’ resulting in a puzzle that does not test and teach the player, reducing the effectiveness of SDG 4 (2012).
Exhibition
Reception at Code & Canvas was also positive overall. Players familiar with The Legend of Zelda series noticed it as a point of inspiration. A diverse range of ages played, younger players generally not understanding aspects of puzzles and older players struggling with combat and movement. Trends mirrored what was found in the playtest. Many people struggled with the double swing, trying to open the boss door with the small key, and not using the 360-degree camera to its full potential.
A Door players wont see if the camera is not facing the correct direction
Strange Swinging Angle
Recreation of someone struggling with the double swing section
Conclusion
This project explored how dungeon design impacts challenge and engagement through pacing, environmental guidance, and mechanical driven progression. Galatea aimed to create a gameplay experience that encouraged exploration and player learning without implicit instruction. Iterative testing proved essential in identifying key issues relating to pacing, puzzles, and the difficulty of certain mechanics, leading to substantial revisions throughout the development process. Changes such as simplifying puzzles, redesigning sections of the level and a general improvement to visual design all contributed to a superior experience for the player.
User feedback also contributed to what players would want to see next. A greater scope would allow for the implementation of boss fights, more items and their practical use inside combat and puzzles. Ultimately giving the project direction for the future, in terms of both a high-fidelity demo and potentially full game that has a concrete understanding of what motivates, engages, and challenges a player.
References
@thomraktar (2026) 6 February. Available at: https://www.instagram.com/p/DUa18spD-Vm/ [Accessed 21 May 2026].
Barr, M. (n.d) Take the Bartle Test. Dr Matthew Barr | Matt Barr’s website. Available at: https://matthewbarr.co.uk/bartle/index.php [Accessed 21 May 2026].
Bartle, R. (1996) Hearts, Clubs, Diamonds, Spades: Players who suit MUDs. Journal of MUD research. Available at: https://mud.co.uk/richard/hcds.htm [Accessed 21 May 2026].
Blender (2026) Blender Manual: Array Modifier. docs.Blender.org. Available at: https://docs.blender.org/manual/en/latest/modeling/modifiers/generate/array.html [Accessed 21 May 2026].
Blender (2026a) Blender Manual: Linked Duplicate. docs.Blender.org. Available at: https://docs.blender.org/manual/en/latest/scene_layout/object/editing/duplicate_linked.html [Accessed 21 May 2026].
Breath of Fire IV (2000) PlayStation [Videogame]. Japan: Capcom.
Larochelle, S. (2014) Depicting the Level Design of a Legend of Zelda Dungeon. Game Developer. October 13. Available at: https://www.gamedeveloper.com/design/depicting-the-level-design-of-a-legend-of-zelda-dungeon [Accessed 21 May 2026].
Markom3D (2019) Blender Level Design Tutorial - Tips. Available at: https://www.youtube.com/watch?v=X0mxcoHVEO0 [Accessed 21 May 2026].
Moffat, D., Crombie, W., Shabalina, O. (2017) Some Video Games Can Increase the Player's Creativity. International Journal of Game Based Learning, 7, (2). Available At: https://go.gale.com/ps/i.do?id=GALE%7CA756288866&v=2.1&it=r&u=ucwinch&p=AONE&aty=shibboleth [Accessed 21 May 2026].
Norman, D. (2013) The Design of Everyday Things. 2nd edn. United States: Basic Books.
NHS (2025) Epilepsy. Available at: https://www.nhs.uk/conditions/epilepsy/ [Accessed 21 May 2026].
Portal 2 (2011) PC [Videogame]. United States: Valve.
Quantic Foundry (2015) The 9 Quantic Gamer Types. Available at: https://quanticfoundry.com/gamer-types/ [Accessed 21 May 2026].
Stout, M. (2012) Learning From The Masters: Level Design In The Legend Of Zelda. Game Developer. January 3. Available at: https://www.gamedeveloper.com/design/learning-from-the-masters-level-design-in-i-the-legend-of-zelda-i- [Accessed 21 May 2026].
The Global Goals (2012) Goal 4: Quality Education. Available at: https://globalgoals.org/goals/4-quality-education/ [Accessed 7 May 2026].
The Legend of Zelda: A Link Between Worlds (2013) 3DS [Videogame]. Japan: Nintendo.
Unity Docs (2026) Set the culling mode in a shader. Available at: https://docs.unity3d.com/6000.3/Documentation/Manual/set-culling-mode.html [Accessed 21 May 2026].
Unity Docs (2026a) Character Controller. Available at: https://docs.unity3d.com/ScriptReference/CharacterController.html [Accessed 21 May 2026].
Unity Docs (2026b) Rigidbody. Available at: https://docs.unity3d.com/ScriptReference/Rigidbody.html [Accessed 21 May 2026].
Unity Docs (2026c) Input System. Available at: https://docs.unity3d.com/Packages/com.unity.inputsystem@1.19/manual/index.html [Accessed 21 May 2026].
Unity Docs (2026d) Camera - Orthographic Size. Available at: https://docs.unity3d.com/ScriptReference/Camera-orthographicSize.html [Accessed 21 May 2026].