research: prioritize actual recommendation hit rate

[choigod1023]

Final objective

Persist higher actual recommended-pick hit rate in AGENTS.md and an always-applied project rule. Auxiliary metrics never substitute for recommendation outcomes. Require denominators, coverage, equal-count comparisons, league/time breakdowns and untouched validation.

Executed experiment

• Fixed market/process agreement ranking committed before scoring; actual frontend eligibility reused.
• Reused K1 cache: 330 games, 2025-2026. Market/research baseline, not reconstructed production predictions.
• Equal one-pick/day budget: 21/42 (50%) to 20/42 (47.62%), same 12.73% coverage. Candidate rejected.
• Full recommendation-count control: unchanged 24/47, identical picks.
• Odds capture times unverified; repeated test period exploratory only. No promotion or improvement claim.
• Integrated rerun reproduced metrics; source SQLite hashes unchanged. Raw picks remain ignored/private.

Verification

768 tests and 58 subtests passed. Markdown reports include failures and limitations.

No scheduler, production changes, merge or deployment. Persistent guidance reaches other main-based tasks after merge.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 3456f749-fea4-4bb7-aed0-d6c8b8b6dfe4

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}