Ralph Loop 2026-03-16: Candidate Review Packet

Change

Added a bounded live candidate-review packet for PRD 003 so the loop now explains current near-miss and deployable rows instead of only reporting aggregate replay/search scores.

Files: - research/candidate_review.py - research/continuous_loop.py - research/status.py - research/discord_status.py - research/README.md

Why

research/latest_state.json and the status commands were still weak at answering:

Which current rows are closest to a trade?
Which sleeve would own them?
Why are near misses blocked?

That made “no trade” opaque even when the current scan already contained sleeve-qualified rows.

Result

The new packet:

loads the latest scan rows from the last 6 runs
classifies each row as research_candidate, review_candidate, or deployable_trade
records the closest sleeve family plus explicit invalidation reasons
exposes top blockers and top candidates in status output

Focused smoke output on current data:

deployable_trade=3
review_candidate=6
research_candidate=27
top deployable: KXTAIWANLVL4-26JUL01 via politics_midprice_reversal
top review blocker: live_price>0.01
dominant blocker mix: heuristic_direction=neutral, live_price<0.7, category!=Economics

Validation

Commands run:

python3 -m py_compile research/candidate_review.py research/continuous_loop.py research/status.py research/discord_status.py research/strategy.py research/search.py
python3 research/candidate_review.py --db data/unicorn.db --recent-runs 6 --max-per-state 3
python3 research/status.py
python3 research/discord_status.py

Observed behavior:

All updated modules compiled.
The candidate-review helper produced structured deployable, review, and research packets.
Human and Discord summaries now surface the candidate-review counts and top invalidation reasons.

Next bottleneck

The packet showed that the current live scan has deployable politics-midprice rows while the stored loop snapshot still reports zero actions. The next bounded change should tighten freshness/handoff semantics so latest_state.json cannot mask currently deployable rows behind stale replay state.