Signal backtest
Do disclosed trades beat the market — and does the edge survive the disclosure lag? For each transaction we measure forward price return vs SPY from two entry points: the trade date (the insider's edge) and the filed date (what a subscriber could actually act on). Alpha = trade return − SPY return over the same window.
Median disclosure lag: 10 days — short enough that most of the edge is still capturable at filing.
Coverage: 473/2672 eligible trades priced (55% of disclosed $). The unpriced 45% are delisted/foreign/unresolved tickers, silently excluded — a survivorship bias that flatters the numbers below (names that went to zero can't report a forward return).
Mean alpha vs SPY (percentage points)
| Side | Basis | 1m | 3m | 6m | 12m | Hit-rate (3m) | n (3m) |
|---|---|---|---|---|---|---|---|
| buys | trade-date | +2.7 | +5.9 | +8.7 | -0.3 | 53% | 203 |
| filed-date | -2.5 | +4.5 | +11.4 | -1.0 | 55% | 196 | |
| sells | trade-date | -4.3 | -3.4 | -6.9 | -10.9 | 31% | 241 |
| filed-date | +2.2 | +0.8 | -4.5 | -8.8 | 41% | 229 |
Buys want positive alpha; sells want negative (they exited before under-performance). Hit-rate = share of buys beating SPY.
By sector (1m)
| Sector | Alpha | n |
|---|---|---|
| Information Technology | +4.1 | 87 |
| Energy | +3.6 | 12 |
| Consumer Discretionary | +0.8 | 26 |
| Health Care | -0.1 | 26 |
| Financials | -0.9 | 19 |
| Utilities | -1.8 | 7 |
| Industrials | -2.3 | 24 |
| Real Estate | -2.5 | 3 |
| Materials | -3.6 | 5 |
| Other | -3.8 | 6 |
By person (1m, ≥5 buys)
| Person | Alpha | Signal | n |
|---|---|---|---|
| Nancy Pelosi | +9.9 | 60 positive | 16 |
| Ro Khanna | +4.2 | 48 neutral | 5 |
| Marjorie Taylor Greene | +1.4 | 80 strong | 68 |
| Josh Gottheimer | +0.8 | 65 strong | 84 |
| Virginia Foxx | -0.9 | 45 neutral | 22 |
| Scott Franklin | -1.1 | 33 lagging | 17 |
Signal = blended 3–12m buy alpha, shrunk toward 50 for small samples.
Method & limitations: forward price return (ex-dividends) over each horizon, equal-weighted per disclosed transaction (not $-weighted). Benchmarked against SPY over the identical window. Delisted/unpriceable tickers drop out (survivorship bias — flatters alpha). Hit-rates near 50% mean positive mean alpha is driven by a right-skewed subset of winners, not a consistent per-trade edge — a portfolio tilt, not a trade signal. A directional study, not a trading model.