Story Lab / Shepherd Maze / Learning record · September 2026
We tested whether a shepherd’s journey to the Nativity could become a satisfying maze: automatic walking, simple remote inputs, junction choices and memory.
Decision: preserve the maze for comparison; develop exploration and discovery in Shepherd Adventure. The playtest exposed a mismatch between the information players could see, the decisions they were asked to make, and the atmosphere we wanted them to experience.
This records why this implementation did not meet our goal. It is not a conclusion that mazes cannot work.
01 / What we were testing
These are representative saved artifacts, not a matched before/after pair or a recording of the exact reported moment. Later builds varied in presentation and map assistance.
02 / The central mismatch
Two visible choices, but no visible reason to prefer one. What would inform your decision?
The behind-the-player view felt restricted by maze walls. It showed surroundings, but not enough route structure to reason comfortably. Map assistance could help; it did not make the in-world view itself more informative.
For another maze: decide whether players solve through an overview, remembered landmarks or local evidence. Make that information reliably readable.
The playtester described it as neither a satisfying pen-and-paper maze nor a character solving interesting challenges in front of them. Walking and remembering turns were not enough to make the middle of the story engaging.
For another maze: define the interesting mental action at a junction. Traversal time alone is not challenge depth.
Basic arrows were an intended TV constraint. Their combination with the fixed follow view made travel feel mechanical and limited opportunities to experience the setting. The constraint was valid; the camera response needed more thought.
Keep the simple input, but design camera composition and movement around it. Test what players see before they commit to a turn.
The playtester wanted a shepherd approaching a small settlement from dark fields. Broad visibility, bright lighting and a plain flat road weakened that feeling. This was a presentation problem alongside the gameplay problem, not proof that a maze needs more decoration.
Use light, material and landmarks to support orientation and mood together. Judge darkness from the playable camera, not only a scene render.
03 / The storytelling consequence
Our design interpretation: the challenge should help the player experience the search, not merely occupy the time between story scenes. The pivot gave village details a purpose: investigate landmarks, interpret clues and discover a way onward.
04 / Later iteration lessons — separate from the original maze
Early replacement choices felt like simple maths, without meaningful risk/reward. Later, The playtester walked every available route but missed ↑ investigation. A puzzle cannot rely on players discovering an unexplained control.
Teach the interaction clearly. Make the first useful discovery memorable, and reserve route reveals for specific places.
The direction-following camera improved engagement, but trees and buildings caused occlusion, jitter and views through foliage. A passable character route is not automatically a usable camera route.
Test turns, stops and obstructions together. Camera clearance and stable occlusion handling belong in the layout review.
05 / What the time bought us