The Whale Confirmation Coach + verification layer.
A self-graded test isn't a test. The framework's whale-sentiment score used to be a closed loop — it would compute its own value from inputs, then narrate that value back to the trader as if it were validated. v6.3 broke that loop. The Whale Confirmation Coach (WCC) adds a verification layer that runs web_search against SEC filings, dark-pool reports, and analyst sentiment to grade the underlying score with one of three verdicts: CONFIRMED, CONFLICT, or WEAK SIGNAL. The verification is the entire point. A score that agrees with itself isn't information; a score graded by an independent sweep of public data is.
The three verdicts
- CONFIRMED — the verification sweep finds independent corroboration. SEC filings show institutional accumulation matching the score. Dark-pool prints and analyst upgrades align. The original whale-sentiment value gets full weight in the audit.
- CONFLICT — the verification sweep finds material disagreement. The score said institutional accumulation; the verification finds insider selling of $435M (the PLTR case study from v6.3 testing). The CONFLICT verdict is often more useful than the score itself — it surfaces the disagreement explicitly so the trader knows the underlying signal is unreliable.
- WEAK SIGNAL — the verification sweep finds neither confirmation nor disagreement. Insufficient data to grade the score; no recent SEC filings, sparse dark-pool prints, no analyst movement. The score gets reduced weight in the audit (50% of CONFIRMED weight).
How the verification sweep runs
For each name on the watchlist, the WCC fires:
- SEC EDGAR query for recent 13F filings, Form 4 insider transactions, 8-K material disclosures (last 30 days)
- Dark-pool ATS report parse for the ticker (last 14 days) — looking for unusual volume concentration off-exchange
- Analyst sentiment sweep — recent rating changes, price-target moves, EPS revision direction
- Cross-check against the underlying whale-sentiment score's directional read
The sweep takes ~3-8 seconds per name and runs as part of the audit cycle. Results are cached for 30 minutes (intra-cycle stable), refreshed on next cycle.
The PLTR case study
From v6.3 development: PLTR's whale-sentiment score read +72 (moderately bullish) based on dark-pool volume concentration patterns. The WCC verification fired and surfaced $435M in insider sales over the same window — Form 4 filings the underlying score couldn't see. CONFLICT verdict. The audit downweighted the +72 to roughly +35 effective; the trader who would have entered on the bullish whale read got a pillar refusal with the explicit reason ("CONFLICT — insider selling contradicts whale signal"). PLTR subsequently dropped 14% over the following 8 sessions.
The case isn't about predicting the drop — the framework doesn't predict. It's about catching the disagreement that the closed-loop score missed. The verification is the value-add.
Why a CONFLICT verdict often beats a CONFIRMED one
Counter-intuitive but consistent in the framework's logs: a CONFIRMED verdict mostly tells you what you already suspected (the score agrees with publicly-readable signals — fine, no surprise). A CONFLICT verdict tells you something you couldn't know without the sweep — the score is wrong, or at least disputed. That's the higher-information event. Most retail traders never run an independent verification because it's expensive in time; the framework runs it on every audit cycle and surfaces it as a single chip.
The operational consequence: don't dismiss CONFLICT. It's the verification working. The trader who takes the trade anyway and overrides the verdict has explicit ownership of the disagreement; the journal records the override.
What the framework does
- WCC chip on every audit card with the verdict (CONFIRMED green / CONFLICT red / WEAK SIGNAL amber)
- Hover tooltip shows the specific verification finding (e.g., "13F filings show no Q3 accumulation; Form 4 shows $52M insider sales")
- Score weight adjustment in the composite audit: 100% on CONFIRMED, 50% on WEAK, ~25% on CONFLICT (effectively neutralizes the score)
- Refresh cadence — every audit cycle refreshes the verification, so a new SEC filing surfaces within hours
The real lesson
A self-graded test isn't a test. The Whale Confirmation Coach's verification layer is the framework's mechanism for not trusting its own scores. CONFIRMED corroborates; CONFLICT disputes; WEAK SIGNAL flags insufficient data. The verification is run automatically on every audit cycle, on every watchlist name. The trader's job is to read the verdict alongside the score — not to ignore it because the underlying number "looked good." A CONFIRMED 60 is more tradeable than a CONFLICT 80. The verification matters more than the score it's grading.
Related: L17 — hidden tape · L31 — AI surfaces