How We Score This
Our methodology is published openly because transparency is the product. Every score is hypothetical and backtested. We show sample sizes because small N means thin estimates. We tell you when a famous signal is noise.
Hypothetical / Backtested Results
All performance numbers on this site are hypothetical. They assume you could buy at the close on disclosure date, ignore slippage, liquidity, and trading costs. Past performance does not indicate future results. This is not investment advice.
Performance Scoring
We ask: “If you had followed this filer from disclosure date, how would you have done?” Returns are measured from when a follower could actually act, not from trade date.
Cumulative Abnormal Return (CAR) over 1w, 1m, 3m, 6m horizons, benchmark-adjusted against SPY.
Win rate — what fraction of trades beat the benchmark over the horizon.
Sample size (N) shown on every score. Minimum 5 trades to display; under 20 flagged as thin estimate.
Multiple-testing correction — across hundreds of filers, some will look good by chance. We account for this.
Signal Classification
Not all trades are equal. We classify by the information content of each filing.
| Tier | Pattern | Why |
|---|---|---|
| Strongest | Cluster of discretionary buys, multiple insiders, short window | Most-cited smart-money signal in academic literature |
| Moderate | Discretionary sells, especially clusters or pre-earnings timing | Sells happen for many reasons; clusters/timing add weight |
| Low | Any transaction flagged 10b5-1 (scheduled) | Decided months ago — low current information |
| Noise | Grants, exercises, tax withholding, gifts (codes A/M/F/G) | Compensation mechanics, not investment decisions |
Bill Passage Model
We estimate passage probability using structural factors only. The model does not use trading data as input — passage odds and positioning are independent layers.
Factors:
- Sponsor is majority-party member
- Bill assigned to a favorable committee
- Number and bipartisanship of cosponsors
- Reintroduction (introduced in prior sessions)
- Whether the bill is a must-pass vehicle (appropriations, NDAA)
- Provisions that may ride into larger packages
Data Sources and Freshness
| Data | Source | Lag |
|---|---|---|
| Corporate insider trades | SEC EDGAR (Form 4) | ~2 business days |
| Congressional trades | House/Senate periodic reports | ~45 days |
| Bill status & metadata | Congress.gov API | Daily |
| Stock prices | EOD market data | Daily close |
What We Do Not Claim
We do not claim that trading positions predict bill passage.
We do not claim that following any signal guarantees returns.
We do not allege illegality. Form 4 and STOCK Act disclosures are legal, public filings.
We do not promise that backtested performance will continue.
The best academic evidence shows a modest edge (~0.70 AUC), not magic. We present it as one input among many.
Questions about our methodology?
Explore the data