← Back to dashboard

Project Closeout Roadmap

Generated 2026-05-15 · Status: in_progress

Diagnostic stack is complete. Closeout means promoting selected diagnostics to either (a) reviewer-actionable state or (b) decommissioning, plus one cycle of manual review. No production scoring formulas were changed in this batch.

Diagnostics shipped (no production score change)

Activity-Adjusted Review diagnostic

Open report

Purpose: Show how the top of the board would shift under a transparent activity overlay (liquidity / rel-vol / Pine accumulation / overextended). Now ships a top-25 entrants/drops section so the reviewer can spot promotion candidates without reading every row.

Promotion criteria: If the entrants list is consistently higher-quality than the current top-25 in manual review across at least 3 sessions, propose folding a small (≤±5%) version of the multiplier into production. Otherwise keep diagnostic-only.

External Benchmark + Disagreement Queue Review review workflow

Open report · CSV export · starter template

Purpose: Surface tickers where internal scoring disagrees with external benchmark sources; carry Pine and cool-off context; suggest keep / watchlist_only / ignore / needs_more_data per row. Now exports a CSV for offline review and a starter state template.

Promotion criteria: After one full review pass (state file no longer all-blank), measure: do reviewer 'ignore' decisions land on names that later confirm the internal view? If yes, keep diagnostic; if not, tune the gap thresholds in external_benchmark_review.py.

Pine Go/No-Go Diagnostic + Cool-off Cohort Tracking overlay

Open Pine report · Open cohort report

Purpose: Daily-OHLCV proxy of the Pine v6 gate stack; A/B cohort tracking of overextended_bb / clean_go forward returns.

Promotion criteria: Already production-quality as a gating overlay. Promotion is gated on the cool-off cohort showing a statistically meaningful spread over enough sessions (target: 20+ trading days).

Accumulation Signal Meter v1 diagnostic

Open report

Purpose: Per-ticker accumulation-style score built from Pine OHLCV metrics (rel-vol, MFI, RSI zone, bar strength, near 20d high, close location). Diagnostic only; surfaces top/weak accumulation names and overlap with Pine clean-go and activity-adjusted top-25.

Promotion criteria: Track agreement between top-25 accumulation list and the eventual reviewer-confirmed names from the disagreement queue. If 60%+ overlap on confirmed-keep decisions over a month, consider adding as a small weighted input to ai_score (would require explicit user signoff).

Manual review steps for the next session

  1. Open the dashboard and click into Activity-adjusted. Compare the top-25 entrants and drops. Note any clearly-wrong activity-adjusted entries (e.g. earnings within 7d, low-quality momentum).
  2. Open Disagreement review. Optionally download the CSV (link at the top) and edit the starter template data/reports/disagreement_review_state_template.csv offline. For each severe row, decide keep / watchlist_only / ignore / needs_more_data.
  3. Persist decisions: edit data/reports/disagreement_review_state.json (the canonical state) to record reviewed=true, decision, and notes for each reviewed entry. The next run will preserve those edits.
  4. Open Accumulation meter. Cross-reference the top-25 with Pine clean-go and activity-adjusted top-25 (overlap sections are inline). Names that appear in all three are the highest-conviction diagnostic candidates.
  5. Open Pine go/no-go and cool-off cohorts. Verify no new overextended_bb blockers on names you decided to keep.

Remaining backlog / known scope decisions

Deploy steps (manual)

  1. Review the diff locally: git log --oneline -5.
  2. (Optional) Re-run reports locally: python 02_Code/Python/Reports/activity_adjusted_review.py && python 02_Code/Python/Reports/disagreement_queue_review.py && python 02_Code/Python/Reports/accumulation_signal_meter.py.
  3. Push to main (the agent will push if the working tree is clean).
  4. Trigger the GitHub workflow manually when convenient — Actions → 'Update rankings' → Run workflow. This will re-generate all reports against fresh data and stamp tasks.json.
  5. After the workflow run, refresh the dashboard; verify the new report links (Activity-adjusted, Accumulation meter, this roadmap) and the new CSV link in the Disagreement review page.