Research program · specification published

Does following a person still work after everyone can see the trade?

Public disclosures can be tested as strategy inputs. Validated records when an action happened, when it became public, and the first realistic follower price. A famous name, a screenshot, or a profitable holding is not a verdict.

See the test protocol Browse equity research

No person-level strategy has been validated in this program yet. Existing insider cohort records must be reproduced from point-in-time filings before they can advance.

Signal lanes

Different people create different data problems.

The sources are not combined into one popularity score. Each lane keeps its own identity standard, reporting delay, and executable test.

01First media profile seeded

Public investors

Primary feed: Podcasts, interviews, newsletters, and dated social posts
Known delay: Public immediately; trade timing and position size may be unknown
First test: Separate thesis mentions from claimed trades, then enter only after publication

02Cohort candidates exist; person-level validation pending

Corporate insiders

Primary feed: SEC Forms 3, 4 and 5
Known delay: Usually up to 2 business days
First test: Open-market purchases by role, size, clustering, and prior buying history

03Source mapping pending

Elected officials

Primary feed: Official periodic transaction reports
Known delay: Disclosure can arrive weeks after the trade
First test: Trade only after the filing becomes public; preserve reported value ranges

04Research design pending

Institutional managers

Primary feed: SEC Form 13F
Known delay: Up to 45 days after quarter end
First test: Position changes, concentration, persistence, and overlap—not assumed entry prices

05Identity and execution controls pending

Public on-chain accounts

Primary feed: Venue and chain events
Known delay: Near real time, identity confidence varies
First test: Executable follower fills after detection, including fees, slippage, and address drift

First tracked investor

A source ledger, not a fan page.

Profiles begin with timestamped claims. They earn stronger labels only when the position evidence and later outcome can be reproduced.

Point-in-time protocol

The public timestamp is the starting gun.

A backtest fails if it trades on the private transaction date when the disclosure arrived later. Every candidate must clear these controls before any performance ranking is shown.

G1Identity

Can the activity be attributed to the same public source without guessing?

G2Availability

What exact time did the information become public and machine-readable?

G3Executability

What price could a follower obtain after publication, with costs and liquidity?

G4Selection

Was the person or cohort chosen before the holdout period began?

G5Persistence

Does the signal survive new periods, different regimes, and source inactivity?

G6Attribution

Is performance distinct from market, sector, size, momentum, and concentration exposure?

Current evidence

Start with the existing insider cohort, then earn the right to name people.

Next executable slice

Ingest raw SEC insider filings with filing timestamps and amendments.
Normalize only open-market purchases; separate grants, options, gifts, and planned sales.
Freeze cohort rules for role, purchase size, clustering, and prior history.
Enter on the next tradable bar after publication and include costs.
Hold out the newest chronological period before ranking any source.

Until that replay is complete, the catalog’s existing insider records remain discovery candidates—not person-level proof and not a recommendation to mirror a trade.