Skip to content

feat: Phase C C9 cheap enemy comparison report -- standardizes anti-hype by requiring all declared enemies to be beaten - #890

Merged
cschanhniem merged 1 commit into
mainfrom
feat/c9-cheap-enemy-comparison
Jul 10, 2026
Merged

cschanhniem merged 1 commit into
mainfrom
feat/c9-cheap-enemy-comparison

Conversation

@cschanhniem

Copy link
Copy Markdown
Collaborator

Implements C9 from NEXT_100_PR_MAP: cheap enemy comparison report.

What

  • cheap_enemy_comparison.py: compute_cheap_enemy_comparison() + format_cheap_enemy_comparison() + CheapEnemyComparisonReport + EnemyResult
  • Advanced scorer earns ranking_authority_granted=True ONLY when it beats ALL declared cheap enemies (strict > or <)
  • Supports higher_is_better=True (precision, AUC) and False (calibration error)
  • Margin computed per enemy; all failed enemies listed explicitly
  • 63 tests; BASELINE 7213→7276

Why

Partial victory against cheap enemies is insufficient — a model that beats charge_threshold but not length_filter can still be mostly a cheap heuristic. All enemies must be beaten before a scorer earns ranking authority. This report makes that standard automatic and auditable.

Disconfirming pass

  • Exact tie does NOT beat an enemy (strict greater/less than).
  • ranking_authority_granted is evidence, not enforcement — pipeline code must check it.
  • No safety/release policy touched.

…ype by requiring all declared enemies to be beaten (#889)
@cschanhniem
cschanhniem merged commit 1cf457a into main Jul 10, 2026
1 check failed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant