Methodology and sources
We state openly which data we take from where and how we process it. Transparency is the most important part of a product like this.
Sources currently in use
- Federal Reserve — FOMC calendar
Meeting dates. The decision is announced on the second day of the meeting at 14:00 New York time.
- ECB — Governing Council calendar
Decision on the second day of the monetary policy meeting at 14:15 Frankfurt time; press conference at 14:45.
- Bank of England — MPC dates and voting history
The Bank Rate decision is announced at 12:00 London time on announcement day; past decision rates come from the MPC voting history file.
- Bank of Japan — MPM schedule and policy statements
The decision is announced on the last day of the meeting as soon as it ends; because the time is not announced in advance, it is shown as a day. The policy rate is read from the target sentence in each statement.
- Reserve Bank of Australia — board schedule and table F1
The decision is announced on the second day of the meeting at 14:30 Sydney time; the cash rate target history comes from RBA statistical table F1.
- Atlanta Fed — Market Probability Tracker
Rate probabilities. Distributions derived from CME three-month SOFR option prices; they refer to three-month average rates, not to individual meetings.
- Bank of England — OIS yield curve
The SONIA OIS forward curve estimated daily by the BoE. The Bank Rate expected after each MPC meeting and the probability of a 25 basis point move at the next meeting are derived approximately from this curve (average forward between meetings).
- ECB — Survey of Monetary Analysts
The analyst survey the ECB runs before each Governing Council meeting: the median and range of the deposit facility rate per meeting, and a probability distribution for the next two meetings. Results are only published as PDF, so they are read from the PDF text.
- CBRT — Survey of Market Participants (EVDS)
Monthly survey; the average expected policy rate for the next three MPC meetings, year-end and 12 months ahead. It reflects the expectations of economists and market participants, not market pricing.
- FRED (St. Louis Fed)
History of the Fed's target range (DFEDTARU / DFEDTARL) — used to fill in decision rates for past meetings.
- BIS — Central Bankers' Speeches
Speech texts. The feed only returns the latest 25 speeches and does not support paging; the archive builds up over time through the daily job. The full text is shown with a licence notice only where the publisher's terms allow it (Fed Board, ECB, BoJ, RBA); for the Bank of England and regional/national banks we link to the source.
- ECB Data Portal (SDMX)
Deposit facility rate — the ECB's current policy rate and past decision rates.
- CBRT EVDS
One-week repo (policy) rate, TP.PY.P02.1H — the CBRT's current rate and past MPC decision rates.
- CBRT — Announcement calendar
The MPC decision is announced on the meeting day at 14:00 Turkey time.
Time zone
Meeting times are read in the source's local time and stored in UTC; the site shows them converted to Turkey time (TRT, UTC+3). In places with daylight saving time (the US, the euro area, the UK) the TRT equivalent can shift by an hour during the year — so we do not assume a fixed offset but convert each meeting individually.
Update frequency
Calendar data is re-fetched from the official pages once a day by a scheduled job. Pages are served with a cache of at most one hour.
Data last changed: 12 October 2026, 21:53 TRT
How rate probabilities are computed
We do not compute the probabilities ourselves. The reason: market-based probabilities require Fed funds or SOFR futures/options prices; these are licensed CME data, and CME's terms of use prohibit automated scraping of its pages. Writing our own calculation engine would not remove this constraint — the input data itself is licensed.
So we show the Market Probability Tracker distributions that the Atlanta Fed publishes publicly, as they are, and state the source clearly. This data may be used for personal and educational purposes only, and the site is positioned accordingly.
An important distinction: these distributions do not belong to a single FOMC meeting but to the average overnight rate over three-month periods. They cannot be compared directly with CME FedWatch-style "probability of a 25 basis point cut at the meeting" figures.
Hawk/dove scoring
Speeches are summarised and scored on a scale from -10 (very dovish) to +10 (very hawkish) by a language model (Claude). The score measures the speech's monetary policy stance; speeches on financial stability or supervision score 0 if they carry no signal.
Three safeguards address the consistency risk: every speech is read with the same fixed prompt template, the output is constrained by a schema (the score is always a number within range) and every step of the scale is defined explicitly in the prompt. Each record stores which model and which prompt version produced it — if the prompt changes, old and new scores can be told apart.
A score of zero can mean two different things: a balanced stance, or a speech that carries no monetary policy signal at all. We separate them — speeches on regulation and supervision are marked "no signal" and are left out of bank averages, which they would otherwise pull artificially towards zero.
Summaries and scores are generated automatically; they have a margin of error and are no substitute for the original text. Each speech page links to the source. BIS publishes only the introductory section of some speeches as HTML; these entries are clearly marked.
Scoring is automated with a language model. Speech texts are collected every day; the summary and score are produced in the same daily job by Claude, with a fixed prompt and scale. The summary is written in Turkish and English at the same time. Because the number of speeches scored per run is limited, new entries may briefly remain unscored; the size of this gap is shown openly on the speech archive and score pages.
Disclaimer
This site is for personal and educational use. The information here is not investment advice. Data is compiled automatically from official sources and may contain errors or delays. Rely on the central banks' own announcements before making any decision.