Research, data analysis, and signal guides from the Signal Congress team.
A free Claude skill that encodes the Signal Congress method: six investigative playbooks, honest score interpretation, and hard claim rules — sequence, never accusation. Validated in three cold-chat evals against live production data. The evals passed. They also caught two real defects in our own data, which is exactly the point.
Signal Congress runs a live MCP server — 18 investigative tools over the same corpus behind this site. Directory listing runs through Claude admin settings that only Team and Enterprise organizations have; we are a one-person shop. The custom-connector path takes two minutes on any paid Claude plan: one URL, one OAuth sign-in.
A single congressional trade is one data point. The signal is convergence — a bipartisan cluster, a federal contract, campaign finance, and lobbying all pointing at one company at once. Of 20,053 trades, only 115 reach three-plus layers. Eight reach the apex.
Across 30,000+ STOCK Act trades, congressional purchases lag the market on average — and high-conviction sells still carry positive 180-day alpha (+4.2%, n=199). A July 2026 recalibration cut the original 64.8% win-rate claim to a coin flip; we updated the post and kept the record.
Most tools show 30,000+ trades with equal weight. The signal is in which ones involve two branches of government pointing at the same company. We built the multi-branch convergence signal to find them — now running on primary-source OGE data covering 1,500+ officials.
The naive answer: congressional purchases LOSE to the market by 0.96% at 60 days — they're worse stock pickers than the index. The real answer is buried in the filter. Cross-chamber bipartisan clusters return +6.22% at 180 days across 503 signals. We show our work.
Between April and May 2025, seven congressional members from both parties quietly bought NVIDIA. Three sat on committees with direct AI/semiconductor jurisdiction. Sixty days later the position was up 36.2% vs. SPY.
A Periodic Transaction Report is a two-page government form that most people misread. We break down every field — transaction date vs. filing date, amount ranges, asset type codes — and explain what actually matters for signal extraction.