1. What we're computing
For each upcoming FOMC (Fed) and ECB Governing Council meeting, we publish the market-implied probability of each possible rate outcome (hold, ±25 bps, ±50 bps, and larger moves when priced). We also publish conditional probabilities, an implied forward rate path, and daily historical snapshots that form a time series.
2. Data sources
- Fed: 30-Day Fed Funds Futures (symbol
ZQ), quoted on CME Globex. Pulled via free market-data providers; fallbacks include Stooq and FRED. - ECB: €STR OIS (Overnight Index Swap) quotes by maturity. Derived implied rates are anchored to the published €STR fixing.
- Meeting calendars: sourced from federalreserve.gov and ecb.europa.eu. Validated annually.
3. Futures → implied rate
The Fed Funds Futures contract settles to the arithmetic average of daily effective Fed Funds rates during its contract month. The CME-standard transformation is:
implied_monthly_average_rate = 100 − contract_price
For a month with a single FOMC meeting on day d of N days:
monthly_avg = (d/N) · r_before + ((N−d)/N) · r_after
Solving for r_after from the observed monthly-average gives the market-implied post-meeting rate.
4. Implied rate → outcome probabilities
We use the CME's standard step-function decomposition. Given a finite set of possible outcomes with post-meeting rates {o₁, o₂, ..., oₖ} and the observed expected post-meeting rate E[r], we distribute probability mass over the two outcomes adjacent to E[r] via linear interpolation, clamped and renormalized:
p_cut = (r_hold − E[r]) / (r_hold − r_cut) p_hold = 1 − p_cut
Three+ outcomes extend the same principle using consecutive adjacent-pair decomposition.
5. Update cadence
- Snapshots run twice per business day: once after the European close and once after the US close.
- Every snapshot carries the timestamp it was computed at, so a number never pretends to be fresher than it is.
- Historical retention: snapshots are kept, and the app charts the last 60 days per meeting.
- What we do not promise is tick-level freshness. These are end-of-session numbers, not a trading feed.
6. Validation
We do not check ourselves against CME FedWatch, because reading that page programmatically is exactly the thing we promised not to do. Validation runs the other way round: the probability math is covered by unit tests on its own, and a diff engine scores what we published before each past meeting against the outcome the committee actually delivered. That scoreboard is the honest measure, since it grades the forecast rather than its agreement with another tool.
7. Known MVP limitations
The scaffold uses a single contract per meeting, which amplifies noise when the meeting falls within 5 days of the month-end. A Phase 2 release will cross-anchor with adjacent contracts. See docs/METHODOLOGY.md §10 for details.
8. Not financial advice
RateRadar shows what the market is pricing. It doesn't predict what central banks will actually decide. Nothing here is a recommendation to trade, invest, or change your financial plans.
The same math runs inside the iPhone app, with the methodology version stamped on every snapshot.
