Event analytics for XAUUSD · CPI & NFP
Second instrument since 08/2026: Nasdaq 100 · narrower basis, same method
Vitus will approve you — usually within a few hours. Drop a quick note in the Circle to speed things up. After that, sign in here with your email and password.
Your details were on the member list — you've been approved right away.
Your email is on the member list. We've sent you a confirmation link — one click and your access is active immediately. The link is valid for 48 hours (check your spam folder too).
The Consumer Price Index and the employment report are the two releases that move the gold price further in minutes than almost any other event. NewsEdge quantifies that reaction: more than five years of US macro data (01/2021 to 09/2026) yield, for each release, a distribution of possible price responses — separated by the magnitude of the surprise, with confidence intervals and case counts stated openly. The individual magnitude classes rest on four to eighteen observations; that is stated next to every figure. For the Nasdaq 100 the same evaluation runs alongside on a narrower basis.
For each release, the conditional reaction distribution of the gold price, resolved by deviation from consensus. Every figure with a confidence interval and case count. Updated the day before and again two hours ahead of publication.
1,112 US macro releases from the evaluation window 01/2021 to 09/2026, tested out-of-sample throughout, transaction costs included in the reported values. The underlying calendar archive reaches back to 2007 (4,696 releases). Configurations without a defensible advantage are reported as such.
Entry window, position scaling and exit criteria for the quarter hour after publication are fixed before the release and reviewable afterwards. Entering twenty seconds ahead of the number carries nothing after costs (45.5 % hit rate, avg −1.5 bp per trade, n = 198); the rule set serves discipline, not an expectation of return.
The members area does not deliver an assessment but an evaluation. For every magnitude class of surprise it states the relative frequency with which the gold price historically reacted upwards, sideways or downwards — each with a confidence interval and the underlying case count.
| Print versus consensus | Bullish | Neutral | Bearish | Median | n |
|---|---|---|---|---|---|
| clearly softer | 7 | ||||
| slightly softer | 5 | ||||
| as expected | 18 | ||||
| slightly hotter | 7 | ||||
| clearly hotter | 14 |
Case counts remain visible. A relative frequency built on five observations carries different weight from one built on eighteen — that distinction belongs in the presentation, not in a footnote.
The figures below come from the same evaluation members see in the terminal ahead of every release — here in the open, without access. Window 01/2021 to 09/2026, measurement window fifteen minutes after publication.
How much of the move can be explained by the deviation from consensus (R²) differs markedly between the instruments: for gold the employment report still carries a noticeable share, for the Nasdaq 100 it has next to no explanatory power — there the core Consumer Price Index carries the most. Anyone trading both markets is working on two different tasks on the same two dates.
R² = share of the fifteen-minute move explained by the surprise. Bar length scaled to 0.40 = full width.
Mean move per standard deviation of surprise. A hotter inflation print went along with just over a quarter of a percent for gold, and with more than twice that for the Nasdaq 100. On the employment report the relationship reverses. The negative sign means throughout: hotter print, falling price.
Figures in basis points per standard deviation. Bar length scaled to 70 bp = full width.
A move counts as quiet below ±17.5 bp for gold and ±14.1 bp for the Nasdaq 100; the threshold comes from the typical quiet range of the same market. On the Consumer Price Index the gold price stayed inside that zone in 21.6 % of cases — 11 of 51 — the Nasdaq in only 9.8 % (5 of 51). On the employment report the two are level: 24.0 % against 25.0 %. Put differently: on the Nasdaq the inflation print almost always forces a decision, on gold every fifth time it does not.
Gold on the employment report, resolved by the magnitude of the deviation. The bars show the share of cases with a rising, unchanged and falling price, to the right the median move and the number of observations.
The first row rests on four observations — it carries no statement and stands there only for completeness; case counts set in red mark exactly that. This distinction is the point: the case count stays visible next to every figure. In the terminal this grid exists for every release and both instruments, with a confidence interval on every share.
Performance here means: how well the reaction to a surprise can be explained, and how clean the data basis for that is. No return is reported — for a reason that appears in figures further down.
| Release | Gold R² | Gold bp per z | n | Nasdaq R² | Nasdaq bp per z | n |
|---|---|---|---|---|---|---|
| CPI m/m | 0.203 | −26.7 | 52 | 0.340 | −59.6 | 52 |
| Core CPI m/m | 0.246 | −27.8 | 52 | 0.379 | −64.2 | 52 |
| Employment report | 0.165 | −34.6 | 51 | 0.068 | −19.1 | 53 |
R² = share of the fifteen-minute move explained by the deviation from consensus. “bp per z” = mean move in basis points per standard deviation of surprise; the negative sign means: hotter print, falling price. Evaluation window 01/2021 to 09/2026.
Evaluation as of: releases up to 09/2026. The price series run current; the historical calendar archive is brought forward at the next reconciliation.
Statistical power comes from the number of observations, not from the complexity of the method. Estimation therefore runs not on CPI and NFP alone but on all US releases sharing the same structure — consensus, surprise, price reaction: producer prices, core PCE, ISM, jobless claims, ADP, JOLTS, retail sales.
Every publication splits into two distinct problems: forecasting the deviation from consensus, and the gold price response to a given deviation. Conflating the two is the most common methodological error in this field. NewsEdge separates them strictly, because only the second holds up statistically.
As an aggregate of professional estimates, the consensus is unbiased by construction. Beating it systematically is the most demanding task in this problem space. Here the model must pass a permutation test against the null hypothesis. Where it does not, the report declares its own directional statement unreliable and makes none.
Here an economically grounded transmission channel exists: the surprise shifts real rate expectations, and gold responds to real rates. The relationship can be estimated across the whole event family and is directional enough within the evaluation window for a scenario calculation. Per target release it rests on roughly fifty evaluated cases; the annual values vary considerably, and the case count is stated next to every figure. This is where the system makes its analytical contribution.
So that validation does not measure what merely looks good in hindsight, four methodological constraints apply:
The model reports the extent to which it draws on each feature category. That weighting is part of the disclosure: it describes the mechanics more precisely than any commentary. Immediate price context is selected most often. This weighting describes where the directional model draws its inputs from — not how well it hits: it fails the permutation test at all three releases (p = 0.29 / 0.35 / 0.84) and therefore makes no directional statement. Differences of a few percentage points are not interpretable at this case count.
The system follows a fixed cadence: two complete computation runs ahead of every publication, automated and without manual initiation.
Calendar, price data, macro series and positioning data are re-sourced and the evaluation is recomputed. The historical calendar archive currently reaches to 09/2026; more recent releases only enter after the next archive reconciliation. From this point the starting position exists in a form that permits preparation with a full day of lead time.
Between the two runs forecasters revise their estimates. That revision is itself information and is reported as a separate position, not quietly overwritten.
At this point nothing is decided, only assigned: the deviation from consensus falls into one of five magnitude classes whose evaluation already exists. The decision has been moved ahead of the moment of publication.
All evaluations refer to this 15-minute window. Price behaviour beyond it is not the subject of the analysis and is not presented as though it had been answered.
This section sits in the main body on purpose, not in an appendix. The standing of a quantitative method is measured in part by how precisely it states its own domain of validity.
Trading leveraged products is high-risk and can lead to a total loss. NewsEdge is an analytics tool, not investment advice and not a trading recommendation — the decision is always yours. In full in the risk disclosure (German).
The objective is not higher trading frequency but a different treatment of the two dates each month on which the gold price moves as far in minutes as it otherwise does across days.
Measure. Contextualise. Decide.
NewsEdge is currently not distributed separately — standalone paid access is in preparation. Until then, access forms part of membership in the REBORN Inner Circle by Trading Phenex, where the personal access codes are issued alongside the daily market work.
Already a member? Just apply or sign in above with your access code.
An analytics dashboard for news trading on gold (XAUUSD) and the Nasdaq 100: ahead of US economic data such as the Consumer Price Index (CPI) and Non-Farm Payrolls (NFP), it shows how the respective market has historically reacted to surprises versus consensus — as probabilities with confidence intervals, evaluated across 1,112 macro releases in the window 01/2021 to 09/2026; the calendar archive reaches back to 2007.
No — deliberately not. The release itself cannot be reliably forecast, and the tool says so openly. What it evaluates is the historical reaction of the gold price to surprises, including the cases where there is no edge at all. Not investment advice.
What is shown is the quality of the reaction models, not a return. That is a deliberate choice: a variant search across 4,912 model configurations per instrument produced no advantage over the simple majority rule on a data segment sealed in advance (gold ±0.0 percentage points, Nasdaq −2.7). What holds up is how systematically the instruments react to surprises — not that a return could be derived from it. Historical evaluations say nothing about the future; not investment advice.
Through membership in the REBORN Inner Circle by Trading Phenex. Members receive their personal access code there and apply with it at the top of this page.
Gold is the benchmark and remains the focus. Since August 2026 the Nasdaq 100 runs as a second instrument under the same procedure, but on a narrower basis: 327 releases across six event types instead of 1,112 across eighteen, and without the futures-market positioning data that does not exist for an index in that form. Case counts appear next to every figure in the terminal. Members switch by toggle between both markets.
Currently nothing extra: the tool is included in the REBORN membership. Standalone paid access is in preparation — no date and no price yet. Until then, the only way in is through the Inner Circle.
For active traders who trade economic releases and want to base their decisions on historical statistics. Important: trading leveraged products is high-risk and can lead to a total loss — please read the risk disclosure (German).