Structure guide
Branching Story Structure: A Writer's Field Guide
Compare branching story structures: time cave, gauntlet, branch-and-bottleneck, loop-and-grow, and open map, with route maps and scope math for CYOA.
How to pick a structure
Estimate your scene budget
Decide how many scenes you can realistically write. A pure tree doubles per level; a bottleneck stays near-linear. The budget rules out structures before taste does.
Name the source of replay value
Decide whether readers return for different endings, a changed hub, or hidden routes. Each motive maps to a different structure.
Match structure to story job
Investigation suits a loop-and-grow hub. A moral gauntlet suits branch-and-bottleneck. A short, vivid one-shot can afford a small time cave.
Draw the route map first
Sketch nodes and edges before prose. If you cannot draw it on one page, the structure is too large to finish.
Why structure is a scope decision, not a style decision
The structure you choose is mostly a decision about how many scenes you will have to write. A pure branching tree, where no paths ever rejoin, doubles its scene count at every choice level: three binary choices is 8 endings, ten is over a thousand. Reconvergence, folding paths back into shared nodes, is the single lever that keeps a branching story finishable. So before debating tone or theme, decide how much your structure multiplies your workload. The comparison table below is sorted by exactly that: how scene count grows as choices accumulate.
Route maps you can draw on one page
Each structure has a recognizable shape. The time cave looks like an expanding fan with no lines ever rejoining. The gauntlet looks like a single thick rope with small loops that always return to the rope; stray off it and you reach an early ending. The branch-and-bottleneck looks like a string of diamonds: paths split, then pinch back to a checkpoint, then split again. The loop-and-grow looks like a wheel: spokes lead out from a central hub and back, but the hub changes as flags accumulate. The open map looks like a region graph where edges are unlocked by state rather than by reading order. If you cannot sketch your project as one of these on a single sheet, it is probably an accidental hybrid that will be hard to finish.
Matching structure to the job the story is doing
Use a time cave only for short, high-variety one-shots where each ending is a punchline and total length stays small. Use a gauntlet when the story is fundamentally linear but you want choices to color the journey and to allow failure; it gives strong authorial control with manageable scope. Use branch-and-bottleneck for the most common case: a story where choices should feel consequential within an act but the plot must reconverge for the next act, like a play with fixed scene breaks. Use loop-and-grow for mysteries, dating sims, and management or relationship stories, where the pleasure is watching a familiar hub change because of accumulated state. Use the open map for exploration and immersive-sim feeling, where the reader assembles the order themselves and gates are unlocked by what they have learned or carry.
The failure edge of each structure
Every structure breaks in a characteristic way. The time cave breaks through combinatorial explosion: writers abandon it once the tree outgrows their stamina. The gauntlet breaks into the illusion of choice, where readers notice every fork snaps back to the rope and stops feeling consequential. The branch-and-bottleneck breaks if the bottlenecks ignore the choices that led into them, making earlier decisions feel erased; the fix is to carry a flag through the checkpoint so the shared scene still acknowledges the path. The loop-and-grow breaks if the hub never visibly changes, turning replays into busywork. The open map breaks through soft-locks, where a reader reaches a state with no valid unlocked edge and gets stranded; the fix is to guarantee at least one always-available path out of every region.
The scope math, made concrete
Numbers make the choice obvious. A pure binary time cave costs 2 to the power of the depth in leaf scenes: depth 3 is 8 endings, depth 6 is 64, depth 10 is 1024. No solo author finishes a 1024-scene project, which is why the time cave is reserved for short pieces. A branch-and-bottleneck with the same 10 decision points but a bottleneck every two levels costs roughly five checkpoint scenes plus a handful of branch scenes per segment, on the order of 30 to 40 scenes total rather than a thousand. A loop-and-grow with one hub and six spokes is about seven scenes plus a few flag-conditional variants, even if the reader experiences far more than seven visits because the hub changes. The lesson is that perceived depth comes from state and reconvergence, not from raw branching. When you feel the urge to add a fork, ask whether a flag on a shared scene would create the same felt consequence at a fraction of the writing cost.
Documenting a structure so collaborators can extend it
A branching project is only maintainable if someone other than the author can read its shape. Keep a one-page node map with stable node names, and a flag table that lists each flag, where it is set, and every scene that reads it. Annotate each edge with its condition, if any, so a reader can see at a glance why a path is or is not available. Mark every terminal node as a designed ending rather than an accident. This documentation is also what makes the project safe to expand: when you add a new region or act, you check it against the flag table to confirm you are reading state that actually exists and not inventing a dependency the rest of the story never sets. Stories that ship and keep growing almost always have this map; stories that stall almost always live only in the author's head.
Quick answers
What is the most common structure for finished interactive fiction?
Branch-and-bottleneck. It lets choices feel consequential within a section while reconverging at checkpoints, which keeps the scene count close to linear instead of doubling at every level.
Why is a pure branching tree so hard to finish?
Because it never reconverges, so scene count doubles at each binary choice. Three choices is 8 endings; ten choices is over a thousand. Without reconvergence or flags, the workload outgrows any single author quickly.
Can I mix structures in one story?
Yes, deliberately. A common combination is a loop-and-grow hub for the middle of the story with a gauntlet finale. The danger is mixing by accident, which produces a graph you can no longer draw or reason about.
How does choose-your-own-adventure differ from these structures?
Choose-your-own-adventure is a reading format, not one structure. Classic gamebooks were often time caves or gauntlets, but a modern CYOA project can use any of these structures underneath the same page-turning experience.