Design branching as state plus selective content divergence, not as an endlessly expanding tree.
Core pattern:
Shared spine → Meaningful choice → Divergence → Consequence → Reconvergence → Remembered state → Later callback
Reconvergence is not inherently fake agency. It becomes fake when the game erases the meaning of the player's choice.
Inputs
Use available:
- story or quest spine
- key decisions
- player intentions
- must-have outcomes
- state variables
- content budget
- VO / cinematic constraints
- required companions, locations, bosses, or endings
- replay goals
Define What Must Truly Diverge
For each choice ask:
- Does location need to differ?
- Does objective need to differ?
- Does relationship state need to differ?
- Does information need to differ?
- Does resource state need to differ?
- Does future interpretation need to differ?
- Does ending outcome need to differ?
Do not create unique content simply because a state variable could represent the consequence more cheaply.
Branching Layers
Use different mechanisms at different costs:
- Line variant — conditional dialogue.
- Reaction variant — same scene, different responses.
- State variant — variables alter future logic.
- Beat variant — one or more unique beats inside shared scene.
- Scene variant — unique scene.
- Quest / mission variant — unique playable content.
- Outcome branch — materially different ending or world state.
Choose the lowest-cost layer that delivers the intended agency.
Reconvergence Types
Hard foldback
Branches rejoin the same event quickly, retaining only state.
Soft foldback
Branches rejoin the same macro plot but with different characters, resources, knowledge, or relationships.
Delayed foldback
Branches remain unique for a meaningful sequence before rejoining.
Periodic braid
Paths repeatedly diverge and reconnect while state accumulates.
Remembered Consequences
When branches reconverge, preserve meaning through:
- relationship variables
- faction reputation
- resources
- companion availability
- conditional dialogue
- callbacks
- altered quest context
- changed information
- changed interpretation of shared scenes
- ending evaluation
Interactive Procedure
- Identify the shared narrative spine.
- Mark decisions that express meaningful player intentions.
- For each decision, define the minimum divergence required.
- Identify safe reconvergence points.
- Define persistent state carried through foldback.
- Add callbacks where players are likely to expect memory.
- Estimate unique-content multiplication.
- Collapse branches that differ only cosmetically at high production cost.
- Check that reconvergence does not erase prior meaning.
- Produce a branch map and state specification.
State Specification
For each decision record:
- variable name
- values
- where set
- where read
- content affected
- reset conditions, if any
Prefer a small number of semantically meaningful variables over dozens of one-off flags when possible.
Output
Shared spine
Decision nodes
For each: decision question and player intentions.
Branch map
Use compact text or Mermaid-style structure when helpful, but do not assume rendering support.
Reconvergence plan
| Branch | Unique content | Reconverges at | State retained | Future callback |
|---|
State model
Production-risk notes
Agency preservation check
Explain what remains meaningfully different after reconvergence.
Failure Modes
Avoid:
- binary trees that multiply indefinitely
- branches that reconverge by forgetting what happened
- one flag per line with no semantic model
- expensive unique scenes for trivial tone differences
- state variables that are set but never read
- late callbacks the player cannot associate with the original choice
- making every choice affect everything
- pretending cosmetic variants are strategic branches
Final Quality Check
Verify that:
- each major divergence corresponds to a meaningful intention or consequence
- reconvergence points are causally believable
- remembered state survives foldback
- future content reads important state
- unique-content cost is visible
- the player receives enough feedback to perceive that choices mattered