Detection Algorithm
Herd Detection scans every STOCK Act disclosure and groups trades by ticker.
A herd fires when 3 or more politicians independently buy the same stock
within a rolling window (configurable, default 180 days back).
Trades involving option exercises, contributions, or donations are filtered out — only
genuine market purchases qualify.
Historical data shows herd trades average −1.8% return (30d).
By the time 3+ politicians have filed, the information edge is often priced in.
Herds signal crowded positioning, not necessarily fresh alpha.
Best use: treat herds as a confirmation layer — if you're already watching a ticker,
congressional convergence adds institutional validation, not a standalone entry signal.
3–4 politicians · moderate
5–7 · crowded
8+ · extreme
Score Calculation — 100-point scale (Q40 + C20 + T24 + V16)
quality_weighted_count = Σ(politician quality ÷ 100).
≥5.0 → 40 pts · ≥3.0 → 30 · ≥2.0 → 18 · else 6
Distinct politicians in the herd.
≥10 → 20 pts · ≥5 → 14 · else 6
Days between first and last trade in window.
≤1d → 24 pts · ≤7d → 18 · ≤14d → 12 · else 6
Combined estimated position value (all politicians).
≥$10M → 16 pts · ≥$5M → 12 · ≥$1M → 8 · else 2
Score = Q + C + T + V, max 100 exactly. 40+20+24+16 = 100. Only herds ≥ 20 are stored.
Tier Thresholds
S
Elite — requires near-max across all 4 layers
≥ 90
A+
High-quality multi-factor convergence
≥ 75
A
Strong signal — quality + count both high
≥ 60
B
Notable herd — current best (MSFT, GOOGL)
≥ 45
C
Baseline — valid but fewer strong factors
20–44
Example path to A tier (60 pts):
qwc ≥ 3 (30) + 5 politicians (14) + ≤7d window (18) = 62 pts → Tier A.
The Quality layer (40 pts max) has the highest leverage — herds of elite traders score
much higher even with moderate counts.