Data products built
from the source.

Filings are read from a third-party transcription of House Clerk and Senate eFD disclosures, not from those systems directly, and its licence is unverified. Upstream keeps a rolling window of the most recent 5,000 trades, so this is not a full history. Filing lateness is derived from the disclosure's own dates. No claim that these trades anticipate returns has been tested.

Retail gasoline and diesel prices come from the EIA directly. The fragility score is a composite this product defines -- spread momentum, diesel acceleration, regional dispersion, level percentile and persistence -- and how those five are weighted is a design choice, not a measurement. The EIA revises published history. Nothing has been validated: no test has been run on whether the score anticipates anything.

One row is one award-modification holding its net obligation. FPDS emits several rows per modification and no field on the transaction route separates them -- all 44 the endpoint accepts were tested against 83 sibling groups and none does -- so funding-line and CLIN detail is not retained. Obligation momentum, concentration and the recompete calendar are derived. Point-in-time reconstruction is valid only from 2026-08-10, after upstream identifier reuse was found to have overwritten 67% of earlier transaction flow. No signal has been tested against returns.

Filings come from the Senate Office of Public Records' own API. Per-filing-year completeness is verified against that API, which is the one thing here checked rather than assumed. Registrant, client, lobbyist and issue rollups are derived from the filings as filed. The API revises history. No claim that lobbying activity anticipates anything has been tested.

The OI/price quadrant describes the trailing window and forecasts nothing. A forward-looking regime state WAS tested against held-out data and failed; it is published as descriptive alongside that result.

Max pain tested over 60 expiries per asset against two nulls — no detectable pull. Net GEX, flip level, zero gamma, the call and put walls and the squeeze score are computed, not validated. Whether dealers are actually long calls and short puts is assumed, not measured — by every GEX product, this one included.

Four of six tiers are built, and all four are DESCRIPTIVE. CYCLE produces coordinates -- growth and inflation momentum -- not states: there is no classification, no regime label and no signal, and at cycle horizon it cannot be backtested, because 25 years holds four to eight cycles. REGIME classifies US dollar liquidity into Expansion, Contraction and Transition over 3,040 days from 2014: its pre-registered test of whether those states precede higher volatility returned NO RESULT, and VIX alone separates forward volatility far better. RESOLUTION classifies BTC perpetual microstructure into twelve states over 2,210 days from 2020-09: its pre-registered test of whether a state's directional claim beats its own raw inputs also returned NO RESULT -- its 5-day hit rate is 50-51% against a coin's 50%, and under one of the two pre-registered weightings it is beaten by the net-flow series alone and by a 5-day mean-reversion rule. REGIME v2 asks the same liquidity question as a FLOW rather than a level -- a coordinate of the level tercile and the 13-week growth tercile, nine cells, 4,615 days from 2007 -- because v1's trailing level score read the 2023-24 liquidity drain as neutral while classifying an identically-paced earlier drain as contraction. v2 has NOT been tested against any outcome: it is published as a description only, v1 stays frozen for its forward test to 2027-10-10, and v2's level axis is known to be a ratchet in four of eight policy eras. None of the four makes a forward claim. Every reading is point-in-time, using only the data that existed on its own date; payrolls were revised in 99.7% of months, median 62k on the monthly change.