05 — Political & Policy Event Data Sources

Purpose: A practical reference for a retail systematic trader (NL-based) who wants to monitor what politicians are saying and doing, and turn that into machine-readable signal that anticipates moves in FX, commodities, equities, and gold. All sources are free or have a free tier, legal, and API-accessible (or RSS at minimum). No scraping required.

Companion to: 03-event-driven-fx.md (which markets move on which events) and 04-positioning-and-flows.md (what positioning to read into the move).

Last updated: May 2026.

1. Politicians' Stock Trades (STOCK Act Disclosures)

The 2012 STOCK Act requires every member of Congress to disclose any transaction over $1,000 (their own or spouse's) within 45 days of the trade. Filings are public on efdsearch.senate.gov and disclosures-clerk.house.gov but the raw filings are awful PDFs — every useful tool below is a parser on top.

Capitol Trades — capitoltrades.com

The most polished UI. Owned by 2iQ Research.
No public REST API — the site has client-side JSON endpoints that change without notice. Avoid scraping; the legal grey area is workable but the bigger risk is silent schema breaks.
Best used for research and back-checking, not pipeline ingestion.

House Stock Watcher / Senate Stock Watcher — housestockwatcher.com, senatestockwatcher.com

Maintained by Timothy Carambat (solo dev, since 2020).
House Stock Watcher S3 bucket has been returning 403 since early 2026. Treat House data from this source as stale; migrate to FMP, Finnhub, or Lambda Finance.
Senate Stock Watcher still operational; data on GitHub: https://github.com/timothycarambat/senate-stock-watcher-data — full JSON dump of every disclosure since March 2020. Updates daily.
Free, no key, no rate limit (it's just a GitHub repo).

Quiver Quantitative — quiverquant.com

Free tier on the website only — visualizations of Pelosi, Tuberville, et al.
API is paid: Hobbyist $30/mo (Tier 1), Trader $75/mo (Tier 1+2), Commercial custom.
Python client: pip install quiverquant.
Best signal-to-cost for a retail trader who wants automated alerts.

Finnhub & Financial Modeling Prep (FMP)

Finnhub — free key, congressional trading endpoint is in the basic tier. https://finnhub.io/docs/api/congressional-trading.
FMP — Senate Trading and House Trading endpoints in the Starter tier (~$15/mo).
Both cover dual-chamber, JSON, real-time after disclosure.

Investable products — NANC and KRUZ

NANC (Unusual Whales Subversive Democratic Trading ETF) — tracks Democrat trades. Top holdings: NVDA, MSFT, AMZN. ~$214M AUM, +8.2% trailing year through Mar 2025.
KRUZ (Subversive Republican Trading ETF) — Republican trades. Top holdings: oil majors (Shell, COP, CVX). ~$54M AUM, +5.6% trailing year.
Holdings rebalance ~weekly off the 45-day-lagged disclosure stream. You can replicate at home for free from the disclosure data — the ETF charges 75 bps for the convenience.

The academic edge

Karadas (2019) — "Did Politicians Use Non-Public Macroeconomic Information in Their Stock Trades?" — finds powerful Senate committee members earned short-term abnormal returns 2004–2010 (pre-STOCK Act). https://www.econstor.eu/bitstream/10419/239672/1/1763179915.pdf
Karadas et al. (2019) and Belmont et al. (2020) both find the abnormal returns largely disappear post-STOCK Act.
Practical takeaway: don't expect to alpha-mine Senator returns directly. The interesting signal is what their disclosure tells you about emerging policy themes (semiconductor concentration, defense buys ahead of a CR vote, etc.).

Python snippet — pull Senator disclosures from the open repo

import requests, pandas as pd

URL = ("https://raw.githubusercontent.com/jeremiak/"
       "senator-filings/main/data/all_transactions.csv")
# Or use Carambat's all_transactions.json:
JSON_URL = ("https://senatestockwatcher.com/api/v2/transactions/all")

df = pd.read_json(JSON_URL)
recent = df[df["transaction_date"] > "2026-04-01"]
print(recent.groupby("ticker")["amount"].sum().sort_values().tail(20))

2. Prediction Markets — Real Money on Political/Policy Events

This is the single highest-signal category for a political-event trader. When Polymarket prices "US strikes Iran by Q3" at 38%, that is real capital with a real settlement, not a poll.

Polymarket — polymarket.com

Crypto-collateralised (USDC on Polygon). By far the deepest liquidity on geopolitical contracts: Iran strike timelines, Fed cut probabilities, election outcomes, sovereign default odds.
Three APIs, all free, no key for reads:
- Gamma API — https://gamma-api.polymarket.com — markets, events, search. ~60 req/min unauthenticated.
- Data API — https://data-api.polymarket.com — positions, trades, holders, leaderboards.
- CLOB API — https://clob.polymarket.com — orderbook, midpoints, price history; trading requires keypair.
Docs: https://docs.polymarket.com.
NL legal note: residents of the Netherlands can trade Polymarket via crypto wallet (Polymarket itself geoblocks US retail, not EU). Stay within standard Dutch tax treatment of crypto holdings.

Kalshi — kalshi.com

CFTC-regulated US prediction exchange. Fiat USD, no crypto.
API free for verified users. Base: https://api.kalshi.com/trade-api/v2. WebSocket + FIX 4.4 also available.
Rate limit: 10 req/sec on most endpoints. Currently 0% fees.
Strong on Fed funds rate decisions, CPI prints, recession calls, and election micro-contracts.
Docs: https://docs.kalshi.com/welcome.
NL access: Kalshi requires US KYC. Not directly accessible to NL residents — read-only data is still public.

PredictIt — predictit.org

US political prediction market, academic exemption from the CFTC, capped at $850/contract.
Free read-only JSON: https://www.predictit.org/api/marketdata/all/. No key.
Liquidity has thinned post-2024 but the API is still up and useful for sentiment cross-checks.

Manifold Markets — manifold.markets

Play money, but with thousands of users and surprisingly tight markets on niche policy questions.
Free public API at https://docs.manifold.markets/api, no key for reads, 500 req/min with key.
Use as a breadth-of-opinion check, not for tradeable signal.

Python snippet — fetch live odds on Trump Iran contracts

import requests

# Polymarket Gamma — search for Iran-related markets
r = requests.get(
    "https://gamma-api.polymarket.com/markets",
    params={"q": "Iran", "active": "true", "closed": "false", "limit": 50},
    timeout=10,
)
for m in r.json():
    last = m.get("outcomePrices", ["", ""])[0]
    print(f"{m['question'][:80]:80}  YES={last}  vol=${m.get('volume24hr', 0):,.0f}")

3. Government Data Feeds (Free Official Sources)

This is where you get the ground truth that the prediction markets are pricing against.

FRED — research.stlouisfed.org/fred

844,000 economic time series from 120 sources, all free.
Register: https://fredaccount.stlouisfed.org/apikeys. Key is a 32-char hex string.
Python: pip install fredapi. ~120 req/min comfortable.
Includes: every Fed funds rate, every Treasury yield, every BLS series, VIXCLS, DXY, oil (DCOILBRENTEU, DCOILWTICO), DJIA, every FX cross.

from fredapi import Fred
fred = Fred(api_key="YOUR_KEY")
ten_y = fred.get_series("DGS10")           # 10-year Treasury yield
brent = fred.get_series("DCOILBRENTEU")    # Brent crude $/bbl
vix   = fred.get_series("VIXCLS")          # VIX close

BEA, BLS

BLS API — https://www.bls.gov/developers/ — free key, employment, CPI, JOLTS. Hard limit 500 daily requests with key, 25 without.
BEA API — https://apps.bea.gov/api/signup/ — GDP, personal income, trade balance.
Both publish on schedule (BLS 08:30 ET monthly) — you can pre-cache and run a diff against consensus to fade or fade-fade the release.

Treasury Direct

Yield curve at https://home.treasury.gov/resource-center/data-chart-center/interest-rates/daily-treasury-rates.csv. No key, just an HTTP GET.
Auction results, debt holdings, TIC flows: https://home.treasury.gov/developer-resources.

Bank of England, ECB, BoJ, BoC

ECB Statistical Data Warehouse — https://data.ecb.europa.eu/help/api/overview. Free, no key, SDMX format. Has every ECB policy rate, TLTRO uptake, balance sheet line.
Bank of England — https://www.bankofengland.co.uk/statistics/data — bulk download, no API key.
BoJ — https://www.stat-search.boj.or.jp/ — clunky but free.
BoC — https://www.bankofcanada.ca/valet/docs — Valet API, no key, JSON/CSV/XML.

World Bank, IMF

World Bank Open Data — https://api.worldbank.org/v2/ — no key, JSON/XML.
IMF IFS — https://datahelp.imf.org/knowledgebase/articles/667681-using-json-restful-web-service — JSON over REST, no key, somewhat brittle.

Why this matters for political trading

A Trump statement on Iran is a shock; the FRED data is the state. The reaction depends on where Fed funds, real yields, and oil are when the shock hits. Cache the state nightly; you only need the streaming for shocks.

4. Political Event Calendars

Free, ranked by usefulness

Investing.com economic calendar — https://www.investing.com/economic-calendar/ — has unofficial JSON endpoint and is the de facto industry standard for FX. Scrape-friendly HTML; no rate-limit issues at sub-minute polling.
ForexFactory — https://www.forexfactory.com/calendar — best categorisation by FX impact (red/orange/yellow). No official API. The andrevlima/economic-calendar-api GitHub project parses it cleanly: https://github.com/andrevlima/economic-calendar-api.
TradingEconomics — https://tradingeconomics.com/api/ — free tier limited to a handful of countries; paid for full. Calendar docs: https://docs.tradingeconomics.com/economic_calendar/snapshot/.
Finnhub economic calendar — https://finnhub.io/docs/api/economic-calendar — included in free tier.

Political / legislative

Whitehouse.gov news — https://www.whitehouse.gov/news/ and /briefings-statements/. Public schedule when published.
Congress.gov API v3 — https://gpo.congress.gov/ — free key from https://api.congress.gov/sign-up/. Bills, votes, members, committee schedules. Best for tracking which bills are up for vote this week.
House and Senate floor schedules — published nightly: https://www.majorityleader.gov/floor and https://www.democrats.senate.gov/floor-schedule.
FOMC calendar — https://www.federalreserve.gov/monetarypolicy/fomccalendars.htm. Static, cache annually.
ECB monetary policy calendar — https://www.ecb.europa.eu/press/calendars/mgcgc/html/index.en.html.

5. Political Statement Monitoring

Whitehouse.gov

News page has an RSS feed at https://www.whitehouse.gov/feed/ (current admin republishes this on each transition).
"Presidential Actions" feed at https://www.whitehouse.gov/presidential-actions/feed/ — executive orders and proclamations.
"Briefings & Statements" at https://www.whitehouse.gov/briefings-statements/feed/.

State Department, DoD, Treasury

State Dept press releases: https://www.state.gov/feed/ — every readout, every sanction designation.
DoD newsroom: https://www.defense.gov/Newsroom/Feeds/.
OFAC (Treasury sanctions): https://ofac.treasury.gov/recent-actions — new SDN list designations move FX (RUB, IRR, sometimes CNY).

Federal Register

Every official US gov action in machine-readable form.
https://www.federalregister.gov/developers/documentation/api/v1. No key needed.
JSON, paginated, free. Filter by presidential_document_type=executive_order for EO firehose.

import requests
r = requests.get(
  "https://www.federalregister.gov/api/v1/documents.json",
  params={
    "conditions[presidential_document_type][]": "executive_order",
    "conditions[publication_date][gte]": "2026-01-01",
    "per_page": 100,
  }, timeout=10
)
for doc in r.json()["results"]:
    print(doc["publication_date"], doc["title"])

C-SPAN

Transcripts via the C-SPAN Video Library: https://www.c-span.org/video/ — no formal API; download MP4s and transcribe locally with faster-whisper.

GovTrack — govtrack.us

Wraps Congress.gov plus the unitedstates/congress scrapers.
Free, no key. Bulk data: https://www.govtrack.us/developers/data.
Better metadata on vote ideology and committee membership than raw Congress.gov.

Trump's posts only: best aggregator is trumpstruth.org — RSS at https://trumpstruth.org/feed, polls every few minutes.
GitHub mirror: https://github.com/stiles/trump-truth-social-archive.
The site uses Truth Social's Mastodon-compatible API plus rotated proxies; you are consuming the output, not scraping.
The American Presidency Project also archives Truth Social posts as official record: https://www.presidency.ucsb.edu/documents/app-attributes/truth-social.

Senators and reps' official feeds

Most members publish RSS on their Senate/House sites. The GovTrack pages link to them.
Twitter/X API is now paid; ignore unless you have $5K/mo budget.

Python snippet — unified statement firehose

import feedparser, datetime as dt
FEEDS = {
  "WH news":      "https://www.whitehouse.gov/feed/",
  "WH actions":   "https://www.whitehouse.gov/presidential-actions/feed/",
  "State Dept":   "https://www.state.gov/feed/",
  "Trump TS":     "https://trumpstruth.org/feed",
  "Fed press":    "https://www.federalreserve.gov/feeds/press_all.xml",
  "ECB press":    "https://www.ecb.europa.eu/rss/press.html",
}
for src, url in FEEDS.items():
    f = feedparser.parse(url)
    for e in f.entries[:5]:
        print(f"[{src}] {e.get('published','')} — {e.title[:90]}")

6. War / Geopolitical Event Tracking

ACLED — acleddata.com

The gold standard for armed-conflict event data. Coded by human analysts within ~1 week of event.
Free with registered account; researcher API key. Bulk CSV + REST.
Covers Africa, MENA, S/SE Asia, Latin America, Europe, increasingly N. America.
Docs: https://acleddata.com/data-export-tool/api-user-guide/.

import requests, os
r = requests.get(
  "https://api.acleddata.com/acled/read",
  params={
    "key": os.getenv("ACLED_KEY"),
    "email": "you@example.com",
    "country": "Iran|Israel|Yemen",
    "event_date": "2026-01-01|2026-12-31",
    "event_date_where": "BETWEEN",
    "limit": 0,
  }, timeout=30
)
data = r.json()["data"]

GDELT — gdeltproject.org

Machine-coded global event database, every 15 min, free, machine-readable.
Higher noise than ACLED (it's NLP off newswires) but massively more comprehensive — 100+ languages, 1979 to now.
DOC 2.0 API for full-text article search: https://blog.gdeltproject.org/gdelt-doc-2-0-api-debuts/.
Best path: query via Google BigQuery (free quota covers a hobbyist). Dataset gdelt-bq.gdeltv2.events.
Python client: pip install gdelt.

CFR Global Conflict Tracker — cfr.org/global-conflict-tracker

Editorial summaries, no API. Useful for context layer behind a quant signal.
Categorises conflicts by US-interest impact (critical / significant / limited).

ISW — understandingwar.org

Daily campaign assessments on Ukraine and Middle East with annotated maps.
No API — RSS at https://www.understandingwar.org/rss.xml.
The maps are PNGs; if you want georeferenced front-line data, scrape the daily assessments and pair with ACLED battle events.

Reuters, AP wire feeds

Reuters: only the consumer RSS feeds are public (https://www.reutersagency.com/feed/), and Reuters keeps tightening these.
AP: https://apnews.com/index.rss and similar by section.
Best wire substitute for free: GDELT DOC 2.0 with sourcecountry:US sourcecollection:reuters,apnews filter.

CrisisWatch — crisisgroup.org/crisiswatch

Monthly conflict risk briefing by ICG. Email subscription, no API; valuable as a monthly calibration check for whether your signal is missing entire regions.

Twitter/X data is essentially paywalled. Reddit's API is rate-limited and increasingly hostile. The free, defensible substitutes:

Wikipedia Pageviews API

https://wikitech.wikimedia.org/wiki/Analytics/AQS/Pageviews — no key, no rate limit, daily and hourly granularity since 2015.
Use case: spike in pageviews for "Strait of Hormuz" or "JCPOA" is a leading attention indicator. Research finds explanatory power for index moves but limited predictive outperformance (Gómez-Martínez 2022).

import requests
r = requests.get(
  "https://wikimedia.org/api/rest_v1/metrics/pageviews/per-article/"
  "en.wikipedia/all-access/all-agents/"
  "Strait_of_Hormuz/daily/20260101/20260525",
  headers={"User-Agent": "pol-event-research/0.1"}, timeout=10
)
print(r.json()["items"][-5:])

Google Trends

No official free API since 2025. pytrends (pip install pytrends) still works against the unofficial endpoint but is brittle and rate-limited.
Use sparingly; expect 429s. Cache aggressively.

AAII Sentiment Survey

US retail individual investor sentiment, weekly. Free chart + CSV at https://www.aaii.com/sentimentsurvey.
Contrarian indicator at extremes.

CNN Fear & Greed Index

https://www.cnn.com/markets/fear-and-greed — composite of 7 indicators. No official API.
Crowdsourced historical: https://github.com/whit3rabbit/fear-greed-data.

CBOE — put/call, VIX term structure

Daily total/equity/index put-call ratios: https://www.cboe.com/us/options/market_statistics/daily/.
VIX term structure (VIX9D, VIX, VIX3M, VIX6M, VIX1Y): https://www.cboe.com/tradable-products/vix/term-structure/. Free, updates every ~15s during the trading day.
VIX itself: FRED series VIXCLS — easiest pipe.
Term-structure shape (VIX3M − VIX) is one of the cleanest geopolitical-stress reads.

8. The "Trump on Iran" Playbook

This is the canonical example because it has the cleanest signature: a single political actor can move oil, USD, gold, and equities in seconds with one Truth Social post.

What typically moves, and in which direction

Statement type	Brent / WTI	USD (DXY)	Gold	S&P 500	Defense stocks (LMT, RTX)
Escalation / strike imminent	+5–13%	+ safe haven bid	+1–3%	−0.5 to −2%	+2–5%
De-escalation / talks resume	−3–10%	mild −	mild −	+0.5–1.5%	−1–3%
New sanctions on Iran	+1–3%	+	flat	flat	mild +
JCPOA-type deal hint	−2–5%	flat	flat	mild +	mild −

Historical precedents — the calibration set

8 May 2018 — JCPOA withdrawal: Brent +3% over the announcement day, USD/IRR collapsed (informal market), Lockheed +1.5%. The market had already priced ~70% probability of withdrawal in the week before.
3 January 2020 — Soleimani strike: Brent +3–4% on the day, gold +1.3% to $1,547, S&P −0.7%, DXY mild safe-haven bid. Markets fully retraced within 5 trading days as Iranian retaliation (Al Asad missiles, no US fatalities) was below escalation threshold.
23 March 2026 — Iran strike postponement (the "$580M short-oil bet"): Brent fell from ~$110 to ~$108 on the announcement; an FT investigation flagged $580M in short oil futures placed 15 minutes before Trump posted the statement. Implication: there is a pre-announcement signal somewhere, possibly Polymarket order flow, possibly a leak. (Axios coverage).

Where to monitor in real time

Source	Latency	What it gives you
Trump's Truth — `trumpstruth.org/feed`	~2 min after post	The statement itself
Whitehouse.gov briefings RSS	~5–15 min	Official follow-up
State Dept RSS	minutes	Sanctions designations
Polymarket "Iran strike by date" markets	seconds	Crowd probability shift
Kalshi Iran contracts (read-only NL)	seconds	US regulated odds
GDELT 15-min update	≤15 min	Confirmed media wave
ACLED	~7 days	Ground-truth event coding

The Polymarket / Kalshi contracts to bookmark

"Will the US strike Iran in 2026?" — perpetual, repriced daily.
"Will Iran block the Strait of Hormuz before [date]?" — quarterly.
"Fed cut at next meeting" — for the second-order rate impact.
Search via Polymarket Gamma API with q=Iran, q=Hormuz, q=Israel.

Strategy notes

Trade the second move, not the first. The post-statement spike is almost always overshooting; the fade-back over 1–5 days is a higher-conviction trade.
Use VIX term structure as filter. If VIX3M − VIX < 0 (backwardation) at the time of the statement, expect bigger and longer-lasting moves. If contango is still 3+ vol points, the market is treating it as noise.
Watch open interest on Brent puts the morning of a known FOMC + Iran cluster. Suspicious flows show up there before the public price moves.

9. Implementation — Building a Political-Event Ingestion Pipeline

Architecture

                    ┌──────────────────────────────┐
                    │   Schedulers (cron / Airflow)│
                    └──────────────────────────────┘
                          │            │            │
            ┌─────────────┘            │            └──────────────┐
            ▼                          ▼                           ▼
  ┌──────────────────┐      ┌──────────────────┐       ┌──────────────────┐
  │   POLLED RSS     │      │   PRED MARKETS   │       │   GOV DATA       │
  │ trumpstruth, WH, │      │ Polymarket Gamma │       │ FRED, BLS, ECB   │
  │ State, Fed, ECB  │      │ Kalshi, PredictIt│       │ Federal Register │
  │ 60s polling      │      │ 30s polling      │       │ Daily at 23:00   │
  └────────┬─────────┘      └────────┬─────────┘       └────────┬─────────┘
           │                         │                          │
           └──────────────┬──────────┴──────────────────────────┘
                          ▼
              ┌────────────────────────┐
              │  Postgres / DuckDB     │
              │  (events, prices,      │
              │  series, contracts)    │
              └───────────┬────────────┘
                          ▼
              ┌────────────────────────┐
              │  Nightly "Political    │
              │  Pulse" brief (LLM     │
              │  over delta of events) │
              └────────────────────────┘

Source-by-source ingestion table

Source	Free?	Endpoint / URL	Cadence	Auth
Polymarket Gamma	Y	`gamma-api.polymarket.com/markets`	30s	None
Polymarket CLOB	Y	`clob.polymarket.com`	WS	Keypair (for trade)
Kalshi	Y	`api.kalshi.com/trade-api/v2`	WS	KYC user
PredictIt	Y	`predictit.org/api/marketdata/all/`	1m	None
Manifold	Y	`manifold.markets/api/v0`	1m	Optional
FRED	Y	`api.stlouisfed.org/fred/`	Daily	Key
Federal Register	Y	`federalregister.gov/api/v1/`	Hourly	None
Congress.gov	Y	`api.congress.gov/v3/`	Hourly	Key
BLS	Y	`api.bls.gov/publicAPI/v2/`	On release	Key
ECB SDW	Y	`data-api.ecb.europa.eu/service/data/`	Daily	None
ACLED	Y (registered)	`api.acleddata.com/acled/read`	Weekly	Key + email
GDELT DOC 2.0	Y	`api.gdeltproject.org/api/v2/doc/doc`	15m	None
GDELT BigQuery	Y (within quota)	`gdelt-bq.gdeltv2.events`	15m	GCP creds
Trump's Truth RSS	Y	`trumpstruth.org/feed`	2m	None
Whitehouse RSS	Y	`whitehouse.gov/feed/`	5m	None
State Dept RSS	Y	`state.gov/feed/`	15m	None
Wikipedia Pageviews	Y	`wikimedia.org/api/rest_v1/metrics/pageviews/`	Daily	UA only
FRED VIXCLS	Y	series id `VIXCLS`	Daily	Key
Senate Stock Watcher	Y	GitHub raw	Daily	None

Daily "political pulse" brief — assembly recipe

Each morning at 06:00 NL time, assemble:

Last 24h prediction-market deltas — Polymarket markets whose YES price moved >5pp on >$10K volume. These are real-money repricings; rank descending by |delta| × volume.
Truth Social + Whitehouse statements — text + cluster on entities (Iran, Fed, China, Russia).
Federal Register new EOs, OFAC new designations — published yesterday.
Calendar events in next 48h — FOMC, ECB, BLS releases (NFP, CPI), Treasury auctions, congressional floor votes.
GDELT tone delta — average tone on theme:WB_2024_ANTI_GOVERNMENT_PROTEST and theme:ARMEDCONFLICT for the last 24h vs trailing 30d.
VIX term structure — current VIX3M − VIX, flag if < +1.0.
Politicians' new disclosures — Senate Stock Watcher repo diff.

LLM the seven inputs into a 300-word brief with two charts (prediction-market deltas, term-structure spread). Total cost: ~$0.02/day with Haiku on the summarisation.

10. Three Sources Worth Wiring FIRST

Ranked by (expected signal × ease of implementation) ÷ cost.

1. Polymarket (Gamma API)

Signal: real money on Iran, Fed cuts, recession, elections. The most informationally efficient public source on political probabilities.
Implementation: 30 lines of Python; no auth; in production in an hour. Stream key markets every 30s into Postgres; alert on >5pp moves.
Why first: it embeds everyone else's signal — including the leaks. The 23 March 2026 Iran short-oil incident shows that pre-announcement information shows up somewhere on the chain.

2. FRED

Signal: every macro series that defines the state in which a political shock lands. 10y, 2y, real yields, VIX, DXY, Brent, WTI, gold — all in one place.
Implementation: fredapi + a cron. One day of work to cache the 30 series you actually need.
Why second: every other signal in the pipeline is interpreted against FRED state. Without it you cannot tell whether Trump's Iran post lands in a contango VIX market (fade it) or a backwardated one (chase it).

3. ACLED + GDELT (paired)

Signal: actual ground-truth conflict events (ACLED) cross-checked against media volume (GDELT). When the two diverge — GDELT spikes but ACLED doesn't confirm — you're seeing narrative without substance, which is exactly the noise that fades within 3 days.
Implementation: ACLED weekly bulk pull + GDELT BigQuery query nightly. ~one day of work, mostly schema mapping.
Why third: this is the de-noiser. Most "war news" is GDELT noise; the small fraction that ACLED confirms is the part that actually moves USD/oil persistently.

Stretch picks for later

Senate Stock Watcher — interesting more for theme-mapping than for direct alpha post-STOCK Act.
Federal Register EO feed — high signal on tariff and sanctions days; otherwise quiet.
Truth Social RSS — clean implementation, but Polymarket gets there faster on big statements.

Sources

Politicians' trades & ETFs

Prediction markets

Government data

Conflict / geopolitics

Statements & calendars

Attention / sentiment

Iran-specific event reaction

Political & policy event data — the legal sources