- game.yaml: `givens` (first or later attempt) and `distinct` levels that are not pooled; the shadow roll
depends on the attempt first. Form submissions show the shadow on 9% of first attempts and 0.3% of later ones.
- Reporting bias: a roll whose bias cannot be measured can borrow another roll's (`borrow: {shadow: meadow}`),
so form submissions inform the shadow again. An assumption, stated in planner/policy.yaml.
- policy.json: targets.<colour>.firstAttempt (this attempt, eventually). Black: about 1.4% on a first attempt,
0.07% on a later one.
- solver.js: advisor options `attempt`, `fix`, sampling "uniform"; assign() and armFor() for the plan.
- planner/experiments.yaml + build_experiments.py -> data/public/experiments.json (progress and status per
experiment, rebuilt from the bot's runs).
- python -m collection cycle: import bot runs, refit, re-plan, deliver the kit; .gitea/workflows/bot-cycle.yml.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014jqCC2LZ4x1WHQ1V57nTsv
core
The shared model. Every other part uses it, so run npm test after any change.
The rules and the game are declarative files in spec/; the Python package atlas/ reads them. JavaScript is
used only where it has to run in a browser or in Ken's bot.
| File | What it does |
|---|---|
spec/form.yaml |
Every question of the community form (how to find it in an export) and every answer (what it is coded as), the answers that count as missing, email redaction, when "left your squadmates" was first asked. |
spec/game.yaml |
The game: every decision and its options in order, every roll and what its odds may depend on, every bonding encounter, the flow of a run from the entrance to the dragons, how form answers map onto it, and the text the bot sees for each option. |
atlas/coding.py |
Codes every form row into a run (python -m atlas.coding). Rule tests: tests/core/test_coding.py. |
atlas/model.py, atlas/solver.py |
The route model: what each coded run reveals about each step, the count tables and shrinkage, the posterior draws, and one interpreter of game.yaml used to solve, evaluate and export the best routes (planner/build_policy.py). Tests: tests/core/test_solver.py, test_game.py. |
solver.js |
JavaScript, for the site and the bot. The route advisor: reads the game and the count tables from data/policy.json and re-solves from any point in a run. tests/core/advisor.test.js checks it gives exactly the Python model's values and choices. |
graph.js |
JavaScript, for the bot. Turns the bot's on-screen choices into form answers, and builds the route graph (analysis/network/build-graph.js). Moves to Python with the collection review. |
The site and Ken's bot load the published copies (https://ekcschiano.github.io/threshing-atlas/solver.js,
graph.js, data/policy.json), so those names on the site must not change.
To support a new form answer, add it to spec/form.yaml. To change a rule, edit atlas/coding.py and add a test case.
To add a choice to the game, add it to spec/game.yaml (and, if the form records it, to atlas/model.py).