Oracle
Market-implied NBP policy expectations distilled into a next-move view. The tab strips term premium from the curve via the ACM model with BRW bias correction, lays the survey path of professional forecasters alongside, and blends the implied move with the lexical tone of the latest MPC minutes into a single gauge. The slider walks every panel back through two decades of month-end history.
Read the short: does the curve know the next MPC move?How to read this page (methodology and glossary)
This page condenses the forecast family of the Curves tab into a single question: what path of NBP rate moves is priced for the next two years, and how does the Council's own language lean against it. The NSS market forwards are dropped on purpose. They embed term premium, so they overstate the priced tightening at every horizon. What remains are the two model-implied risk-free paths, the survey path, and the text signal from the MPC minutes. Use the slider to walk the monthly history, every panel re-renders as of the selected date and the playhead on the minutes chart marks where you are.
Glossary
| ACM rf | Risk-free expected short-rate path from the Adrian-Crump-Moench five-factor affine term-structure model, converted to 1-month forwards via the integro-spline construction shared with the Curves tab. |
| BRW rf | Same path with the Bauer-Rudebusch-Wu small-sample bias correction applied to the VAR persistence. The canonical expectations leg on this page. |
| PF | Implied path from the most recent NBP Survey of Professional Forecasters at the snapshot date, time-shifted from the survey date to that date. |
| Δbp | Cumulative change of the 1-month forward at the given horizon versus the NBP reference rate (ACT/365) on the snapshot date. |
| Moves priced | BRW Δbp divided by 25, the standard MPC step size. |
| Net tone | (hawkish minus dovish) over (hawkish plus dovish) word counts per minutes, custom MPC lexicon from the minutes pipeline. Positive readings are hawkish. |
| hike*/cut* | Literal frequencies of the word families hike* and cut* per 1,000 words of each minutes document. |
| Oracle score | 0.70 × market signal + 0.30 × tone signal, shown on a [-100, +100] scale and mapped linearly to P(cut), P(hold), P(hike) over the next quarter. The headline is a stance word, not a move call: LOOSENING below -40, TIGHTENING above +40, NEUTRAL in between, so the gauge cannot contradict the rounded move tiles. Before the first published minutes (2007) the tone leg is treated as neutral. The components are printed inside the gauge tile. |
| Move tiles | BRW implied moves rounded to 25 bp steps with a conservative visibility threshold: a step only registers once the priced move clears 0.75 of it (0.77 moves shows one step, 1.2 still one, 1.75 two). The 6m tile shows only the additional steps beyond 3m, so a flat path past 3m reads "none". The raw bp figure and the exact step count sit underneath. |
What each panel shows
Tiles. The 3m and 6m implied moves in rounded 25 bp steps (the 6m tile shows additional steps beyond 3m), the MPC minutes tone index drawn with the same bar design as the gauge on the plus or minus 100 scale, and the blended next-quarter Oracle gauge with its components.
Implied moves table. The three paths read off at fixed horizons, expressed as forward levels, cumulative moves in bp and 25bp steps priced.
Expected policy path. ACM, BRW and PF 1-month forwards on a 2.5-year horizon with the NBP reference rate anchoring the origin. The gap between BRW and PF is the disagreement between the market-implied model path and the survey at the snapshot date.
History charts. Three panels on one shared timescale with a common playhead at the slider date: net tone per meeting (recomputed from the minutes text extracts at build time, so the latest published minutes are always included), the literal hike*/cut* word frequencies, and the Oracle score itself with its stance bands.
Implied moves vs reference rate
Cumulative change of the 1m forward vs the NBP reference rate (ACT/365) on the snapshot date. Moves priced = BRW change divided by 25 bp.
The three expected-rate paths read off at eight horizons from three months to two years, measured against the NBP reference rate on the snapshot date. The first three columns give the 1-month forward level under ACM, its BRW bias-corrected variant, and the forecaster survey. The two Δbp columns convert the ACM and BRW levels into a cumulative change versus the reference rate, positive for priced tightening. The final column expresses the BRW change in 25 bp steps, the same quantity the move tiles and gauge use.
Expected policy path, next 2 years
ACM risk-free and BRW bias-corrected 1-month forwards, smoothed by an integro-spline construction. PF is the NBP Survey of Professional Forecasters implied path, time-shifted to the snapshot date. NSS market forwards are omitted on purpose, they embed term premium and overstate the priced tightening at every horizon. The dotted grey staircase is the expected policy rate stepped in 25 bp clips from the rounded BRW moves, the discrete read of the same path.
MPC minutes lex gauge
Net tone = (hawkish - dovish) / (hawkish + dovish) word counts per minutes, custom MPC lexicon from the minutes pipeline. Full history shown, the vertical playhead marks the slider date.
Literal hike* and cut* frequencies
Occurrences of the word families hike* and cut* per 1k words of each minutes document. Same timescale and playhead as the lex gauge above.
Oracle index history
The blended score on the [-100, +100] scale at every month-end snapshot, with the NBP reference rate stepped on the right axis. Shaded bands mark the stance zones: LOOSENING below -40, TIGHTENING above +40, NEUTRAL in between. Same timescale and playhead as the charts above.
Disclaimer. The Oracle index, the underlying models and every figure and table on this page are published solely for academic purposes and general knowledge. Nothing here constitutes investment advice, an investment recommendation, a solicitation or an offer to buy or sell any financial instrument, or a forecast on which any investment decision should be based. The index is a research construct built from model-implied expectations and text statistics. It can be wrong, it has been wrong in the past, and its historical performance is no indication of future readings or of future monetary-policy decisions. The author is not affiliated with Narodowy Bank Polski or its Monetary Policy Council, and no content on this page reflects their views. Anyone considering investment decisions should consult a licensed financial advisor. Use of this page is entirely at the reader's own risk.