FX News

Market-moving events, session conditions, and macro context — written for traders who run systems, not predictions.

Follow pairs by starring them on the Signals page.
Tag: uk cpi Clear

Yesterday · 1 notes

We Won Two Legs of Three on Retail Sales. The One That Won Was Scored Against a Prior the Same Release Revised Away.

We pre-registered three losable legs on United States October retail sales and said all three could go against us.

Two won, and the one that won most convincingly won against a prior the same release revised away underneath it. Leg one: the headline landed at plus 0.2 per cent, dead centre of a minus 0.3 to plus 0.7 band, on total sales of 748.2 billion dollars. Leg two: the control group printed plus 0.6 against a threshold of plus 0.2 — but the leg's stated rationale was "at least double its plus 0.1 prior", and one reader now carries that September prior revised to plus 0.53 inside this very release. Against the stored prior the month's rise is 0.48 points; against the revised one it is 0.05, from an identical print — two readings of one month's change in one series, a factor of 9.6 apart. Doubling needed plus 1.06; the print is plus 0.58, a ratio of 1.094. The threshold passed and the sentence justifying it failed. The revision to the input is 3.9 times the entire surprise in the thing we were measuring. Leg three lost cleanly: the Census Bureau says the August-to-September change was unrevised from up 0.8 per cent, and the headline is the field we named, even though the same reader has September at plus 0.83 against plus 0.77 at two decimals, a move invisible at the issuer's own precision and inside its plus or minus 0.2 margin. On that account the September prior of three of four fields moved and the one that did not is the one we built a test around. The control group's absent forecast has also filled, and the way it filled is the finding: two publishers now carry a plus 0.3 consensus for it, both published after the print, neither naming a source. A forecast that first becomes visible after the number it forecasts is not an error bar, it is a reconstruction, and a scraper cannot tell the two apart. Japan's trade balance at 23:50 UTC is the mirror image — one page, a consensus of minus 956 billion yen and a house model of minus 450, 506 billion apart and disagreeing about the direction of the change rather than its size, against a minus 536.6 billion prior. Then the two things that indict the consensus apparatus: the issuer puts a margin of error of plus or minus 0.4 on a release whose surprise was one tenth of a point, four times, and states that the confidence interval includes zero, so the agency will not tell you the sign of the month. Our own half-width of 0.5, drawn from a sample of one, landed 0.1 from that sampling error by an unrelated route, which is luck and we call it luck. And agreement was anti-correlated with accuracy: ex-autos, where two readers agreed to the decimal at plus 0.4, printed plus 0.6 and missed by twice the headline's miss. Vendor agreement measures how much vendors read each other and says nothing about sampling noise. Publishers that stored plus 0.1 call plus 0.2 a beat; the one implying plus 0.2 calls it in line — the same number is a beat or a non-event depending on which pre-release figure a publisher happened to keep.

Trade the reaction, not the headline
News tells you when conditions change. Pair it with the economic calendar and check the current signals before adjusting anything.