Skip to main content
Tools: get_macro, and get_market_regime for the verdict computed from it. Three streams that answer the same question from different directions: how much money is in the system, and what is it currently paying for.

Federal Reserve series

25 series, 79,523 daily rows, from 2010-01-01, synced from FRED.
Each series is shifted by its own publication lag. The Fed’s balance sheet is published weekly and days behind; the policy rate is daily. Net liquidity is Fed assets less the Treasury account less reverse repo: three series with three different lags, so subtracting today’s published values would compare numbers that describe different days.The arithmetic lives in one place, and get_macro returns the inputs rather than the verdict so the shift is inspectable rather than assumed.

On-chain valuation

Daily price, market capitalisation and realised capitalisation for BTC from 2010-07-18 (5,899 days) and ETH from 2015-08-08 (4,052 days). Realised cap values every coin at the price it last moved at, rather than at today’s price. The ratio between the two, MVRV, is what holders collectively paid against what the market currently says, and scoring it against its own history is one of the two inputs to the slow signal.

Stablecoin supply

Circulating supply per stablecoin, daily, USDT from 2017-11-29 (3,209 days) and eight others behind it. Latest reading, 2026-09-11: Stablecoin supply is the closest thing on-chain has to a money supply: it is capital that has already been converted and is sitting in the system waiting. It is kept per issuer and never only as a total, because the total can hold steady while the composition moves entirely, which is itself the reading.

Prediction markets

Polymarket, snapshotted daily: a price ladder per asset per month, and the odds on each upcoming Fed meeting. The snapshot of 2026-09-12 holds 121 ladder rungs across BTC and ETH over four months, and four Fed meetings. Read by get_forecast, where the ladder sits beside the rate curve it is implicitly about.
These are snapshots with an as_of date, not a continuous history: a rung appears when the market lists it and stops when the month resolves, so counting rows across dates measures how often the sync ran rather than how deep the data goes. Read the latest snapshot, which every answer stamps.A prediction market’s odds are only meaningful with their volume beside them, so volume and liquidity are stored per rung and passed through to every answer.

Where these streams stop