Game-rule checks for runs answered through "Report a run" #37

Merged
kathleen merged 1 commits from data/report-rule-checks into main 2026-11-19 14:10:09 +00:00
Owner

Follow-up to #18. Runs answered through "Report a run" (reported and reddit comment) get the game-rule checks the community form already gets in core/atlas/coding.py. These runs are self-submitted, and the Reddit ones are read from free text, so they can be misread or misremembered.

The checks

Each broken rule becomes a data_flags entry; the run is kept.

  • First dragon: outside the documented pool for the company and direction (sighting_pool, same wording as the form's flag).
  • Encounter dragon: a bond or burn by a dragon the encounter doesn't have. This uses a new encounter_colours table in core/spec/form.yaml, next to sighting_pool. Orange/Brown after the two Greens is allowed, as a Green's rescue.
  • Valley Green: a Green on a valley run (same wording as the form's flag).
  • Cadet path: an outcome off its scream path (e.g. a bully outcome after ignoring the scream); fight answers or a post-fight path on a run that didn't help; a second fight move without pausing to assess.
  • Early death: died before the hour, yet reached the tree, the valley or an encounter.

Route model

planner/policy.yaml gains drop_flagged:

  • Flagged runs of reddit comment stay out of the model.
  • The "read from a Reddit comment" provenance note doesn't count as a flag.
  • A hand fix in collection/manual/reddit_comments.yaml that clears the flags brings a run back.
  • There's no effect until reddit comment is in sources. reported runs are not dropped, so policy.json is unchanged.

On real data

On the first 82 scraped runs, 6 are flagged:

  • Brown or Red at the shadow;
  • Green after jumping on the Red;
  • Red in the meadow;
  • Orange first for Alone/Center;
  • cadet-path mixes.

Checks

python -m collection build --check (runs_clean.csv unchanged), experiments/black/tree --check, 89 Python tests, 17 JS tests, graph.json unchanged, the page builds.

🤖 Generated with Claude Code

Follow-up to #18. Runs answered through "Report a run" (`reported` and `reddit comment`) get the game-rule checks the community form already gets in `core/atlas/coding.py`. These runs are self-submitted, and the Reddit ones are read from free text, so they can be misread or misremembered. ## The checks Each broken rule becomes a `data_flags` entry; the run is kept. - **First dragon:** outside the documented pool for the company and direction (`sighting_pool`, same wording as the form's flag). - **Encounter dragon:** a bond or burn by a dragon the encounter doesn't have. This uses a new `encounter_colours` table in `core/spec/form.yaml`, next to `sighting_pool`. Orange/Brown after the two Greens is allowed, as a Green's rescue. - **Valley Green:** a Green on a valley run (same wording as the form's flag). - **Cadet path:** an outcome off its scream path (e.g. a bully outcome after ignoring the scream); fight answers or a post-fight path on a run that didn't help; a second fight move without pausing to assess. - **Early death:** died before the hour, yet reached the tree, the valley or an encounter. ## Route model `planner/policy.yaml` gains `drop_flagged`: - Flagged runs of `reddit comment` stay out of the model. - The "read from a Reddit comment" provenance note doesn't count as a flag. - A hand fix in `collection/manual/reddit_comments.yaml` that clears the flags brings a run back. - There's no effect until `reddit comment` is in `sources`. `reported` runs are not dropped, so `policy.json` is unchanged. ## On real data On the first 82 scraped runs, 6 are flagged: - Brown or Red at the shadow; - Green after jumping on the Red; - Red in the meadow; - Orange first for Alone/Center; - cadet-path mixes. ## Checks `python -m collection build --check` (`runs_clean.csv` unchanged), experiments/black/tree `--check`, 89 Python tests, 17 JS tests, graph.json unchanged, the page builds. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
kschiano added 1 commit 2026-11-19 13:48:20 +00:00
Game-rule checks for runs answered through "Report a run"
CI / test (pull_request) Successful in 2m0s
0f4430f6f5
Reported and Reddit-comment runs now get the checks the community form
gets in core/atlas/coding.py, as data_flags (the run is kept):
- first dragon outside the documented pool for the company/direction
- a bond or burn by a dragon the encounter does not have (new
  encounter_colours table in core/spec/form.yaml; Orange/Brown after the
  two Greens allowed, as a Green's rescue)
- a Green on a valley run
- a cadet outcome off its scream path, fight answers or a post-fight path
  on a run that did not help, a second fight move without assessing
- died before the hour, yet reached the tree/valley or an encounter

planner/policy.yaml drop_flagged: flagged runs of sources read from free
text (reddit comment) stay out of the route model; provenance notes do not
count as flags. No effect today: the source is not in `sources` yet.

On the first 82 scraped runs: 6 flagged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kathleen merged commit fdb603f5a4 into main 2026-11-19 14:10:09 +00:00
Sign in to join this conversation.
No Reviewers
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: DragonKind/threshing-atlas#37