validation/congress_prescreen.py · feed data/congress.py · panel 501 names · entry next-session close, exit +21d · IS<2016/OOS>=2016
| side | n | IS excess | OOS excess | OOS net10 | OOS hit | boot p | placebo p |
|---|
| Buy | 7112 | +0.05% | -0.16% | -0.36% | 47% | 0.895 | 0.995 |
| Sell | 4997 | -0.29% | -0.23% | -0.43% | 47% | 0.952 | 0.985 |
| check | result | detail |
|---|
| A alignment (PIT next-session entry) | PASS | engineered + guard |
| B incrementality (excess vs EW panel) | FAIL | OOS excess -0.16% |
| C economics (OOS net@10 > 0) | FAIL | net10 -0.36% |
| D stability (IS same sign & boot p<0.05) | FAIL | IS +0.05%, boot p 0.895 |
| E null+power (n>=50 & placebo p<0.05) | FAIL | n 7112, placebo p 0.995 |
PRIMARY (Buy) verdict: REJECT / near-miss
| year | events |
|---|
| 2013 | 5 |
| 2014 | 204 |
| 2015 | 416 |
| 2016 | 389 |
| 2017 | 490 |
| 2018 | 530 |
| 2019 | 670 |
| 2020 | 946 |
| 2021 | 847 |
| 2022 | 375 |
| 2023 | 408 |
| 2024 | 375 |
| 2025 | 1114 |
| 2026 | 343 |
| series | n months | total ret | Sharpe | maxDD |
|---|
| copy gross | 240 | +566.9% | 0.82 | -25.62% |
| copy net@10 | 240 | +480.7% | 0.76 | -27.09% |
| SPY B&H | 240 | +753.9% | 0.78 | -50.78% |
- Feed is a structured mirror of official House Clerk PTRs; the PIT timestamp is
disclosure_date (median lag ~10d, p90 ~33d).
- No DSR yet: a full study would add K over the event/exit variants.
- Source linearity: an event study, not a daily strategy; no turnover compounding.