A few trade rows are not enough. Radar now tests meaningful manager expansions, broad repeat buying and cross-source clusters across all four published feeds.
Four ways to qualify
1. Active-manager convergence
At least two different active-manager groups each newly report a position or raise reported shares by at least 10%, increase its portfolio allocation by at least 0.5 percentage points, and end with at least 1% of reported long-share value in it. Qualifying expanding groups must outnumber groups reducing quantity at least 10% from a prior allocation of at least 1%.
This is a quarterly holdings signal, not a confirmed purchase. Raw manager size, passive ownership and fame do not qualify a name.
2. Cross-source purchase cluster
Eligible congressional and corporate-insider purchases must both be present: at least three distinct reporting units, five filer/date buying events, one repeat buyer and a purchase filed within 30 days. Buying actors and events must both outnumber selling actors and events.
3. Repeated insider cluster
At least three reporting units, five filer/date purchase events and two repeat buyers in 90 days, with a purchase filed within 30 days. Buying actors and events must outnumber selling. Known account substitutions, gifts, dividend reinvestment, employee purchase-plan mechanics and purchases in filings marked with a trading-plan indicator are excluded from purchase conviction. Ordinary-share sales remain opposing evidence even when a filing has shared mechanical footnotes or a plan flag. All remain searchable in the source archive.
4. Broad congressional buying
At least five distinct filers, seven filer/date buying events and two repeat buyers in 90 days, with a purchase filed within 30 days. Buying actors and events must outnumber selling. House and Senate remain distinct source feeds, but a few split rows do not create a broad cluster.
What “allocation” means
A security’s reported value divided by the filing’s total value of long-share rows (quantity type SH), excluding put/call positions. It is not a percentage of the manager’s complete assets, net exposure or personal wealth. Different class, option and CUSIP identities are preserved. Multiple rows for the same manager/security are aggregated before comparison, not counted as independent managers.
Only consecutive quarters for the same CIK, full holdings reports, reconciled totals, and non-amended/non-confidential reports qualify. We do not bridge Pershing entities or infer that a missing report is zero holdings. The latest market-wide collected quarter must be no more than 150 days old, and its filing no more than 120 days old.
Share quantities are not split-adjusted; changes may reflect splits, IPOs, transfers or reporting differences. A price move can change a portfolio weight. Requiring both quantity and allocation growth makes the evidence more substantive, but does not prove an executed purchase or motive.
The active-manager cohort
This is an explicit research cohort, not a wealth ranking. Related Pershing reporting entities count as one group. A manager without comparable current quarters can still be searched on the funds page, but does not receive invented position changes.
Berkshire Hathaway
Brad Gerstner · Altimeter
David Tepper · Appaloosa
Daniel Loeb · Third Point
Chris Hohn · TCI
Stanley Druckenmiller · Duquesne
Bill Ackman · Pershing Square Capital Management
Bill Ackman · Pershing Square Inc.
Coatue
Tiger Global
Bridgewater
Renaissance Technologies
Vanguard, State Street, FMR and large bank reporting portfolios remain searchable context, not automatic conviction votes simply because they own many companies.
Source clocks, identity and opposing evidence
Congress and insider conviction uses transaction dates 2026-08-10–2026-11-07; older source rows appear only as clearly dated background. Joint insider owners count as one reporting unit, and same-day split rows count as one event. Different reporting units still do not prove independent investment decisions.
Fund comparisons use the separately labeled 2026-06-30 quarter. Their CUSIPs are joined across filings exactly. Reviewed CUSIP/share-class/ticker/CIK links are required before attaching company-level congressional or insider evidence. Unlinked fund candidates remain eligible under their source name and CUSIP; they are not silently discarded. Multiple linked classes of one issuer produce one company card, not padded stock picks.
All sales, meaningful manager reductions, exact prior/current weights, quantities, filing dates and original sources remain inspectable. The four feeds are evaluated, not assumed to confirm each other. Incomplete coverage never proves no trading occurred.
Ranking and scope
Order: manager groups expanding an already reported positive position, then all meaningful expanding groups, 90-day buying filers, repeat buyers, newest supporting filing, and stable issuer key. First-reported exposure is labeled separately because an IPO or reporting change can create it without new buying. Up to 12 companies are displayed; there is no quota to fill and no hidden performance score. Substantial expansions with an equal or larger number of meaningful reducing groups appear separately as divided signals—not as clear accumulation candidates. Thresholds are product rules, not a statistically validated predictive model.
Published inputs: 3,023 House records, 555 Senate records, 3,090 insider purchase/sale rows and 292,522 institutional position rows. 3,502 fund securities across 9 comparable active managers were screened. The full SEC identity directory replaces the old 67-company sector-only gate.
These are research candidates, not buy recommendations or claims about personal profit. No return badge or predictive claim is made without a separate frozen-rule, point-in-time test including every signal, losses, costs and benchmarks.
