Notícias FX

The Issuer Revised Four of Our Own Priors Inside a Day. And the ECB Tells You in Advance Which Speaker Leaves a Text.

On Thursday this desk published four figures as the Labor Department's own settled prior values for jobless claims, read off the issuer's own release.

Twenty-four hours later the issuer revised all four: initial claims for the week ending 31 October from 189,000 up to 191,000, the four-week average from 194,750 up to 195,250, continuing claims from 1,718,000 down to 1,717,000, and the insured four-week average from 1,709,500 down to 1,709,250. Two up and two down in one document, for the second consecutive week. An issuer is authoritative about its figure as of publication and not about that figure a week later. The cost is in the change field: the new print of 200,000 is a rise of 11,000 against our stored prior and 9,000 against the issuer's revised one, so a stale prior overstates the weekly change by 22.2 per cent of the issuer's own number, and nothing in the release flags it. One field is reported unchanged against an average revised upward by 500 in the same release. The buried figure nobody prints: the unadjusted series rose 28,870 where seasonal factors expected 19,663, an excess of 9,207 against a seasonally adjusted change of plus 9,000 — the headline is the seasonal miss to within 207. Overnight four Fed officials spoke, all hawkish, none naming a size or a date, and Hammack argued explicitly against a half point in December while Waller's 0.3 core CPI condition has already fired. We can source no December probability at all: four pages, no FedWatch figure, and the one number we found reaches us through a translation layer and contradicts our own arithmetic, so we publish the gap. The ten-year is 5.12 per cent on a wrap stamped 21:03 UTC while the free official daily series still shows 4.98 for 10 November — two days and 14 basis points behind. Dollar-yen's exact high was 160.049 and its low 159.299, closing an absence we printed in bold last night and scoring our own implied 159.30 to a tenth of a pip. And the ECB's own calendar prints, in advance, that Elderson at 06:50 UTC will leave no text while Schnabel at 07:20 will — the first time we have found that field before the event rather than after.

December Is Already Fully Priced, So Six Fed Speakers Can Only Move a Ten Per Cent Tail. And the Issuer Says 16:15, Not 16:30.

A quarter-point December hike is fully discounted on the one feed anybody names, which means six Federal Reserve speakers between 14:00 and 23:30 UTC cannot move the base case at all — only a half-point tail priced near ten per cent, and a January meeting nobody has priced. The wire carrying that schedule prints three December figures from CME FedWatch, and they are one observation in three dresses: 25 basis points certain plus a ten per cent chance of a further 25 gives 27.50, the page says 27.6, and solving backwards gives a half-point probability of exactly 10.40 per cent — the same figure we published yesterday from a different route. Store all three as separate inputs and your system counts one number three times. Waller's half-point condition has already fired: he said in London on 10 November that it may be needed if October core CPI came in near 0.3 per cent or higher, and core CPI printed 0.3. The market pays about ten per cent for it. We also settle a dispute we published this morning, and nobody who agreed was right: two aggregators put Goolsbee at 16:30, one put him at 16:15, and the Chicago Fed's own page says 10:15 am CT, which with daylight time over is 16:15 UTC. The lone dissenter matched the issuer. The same page calls the event a moderated virtual discussion with a named journalist, where one aggregator calls it a speech — a scheduled event with no document to read. And the day's only US government release is framed by calendars that are broken in three adjacent rows: three publishers give consensuses of 203,000, 191,000 and 195,000, a 12,000 spread, while the Labor Department's own 5 November release shows 189,000 belongs to the week ending 31 October and 1,718,000 to the week ending 24 October, both filed a week forward, and a four-week average sitting in a level field. The one-week lag between initial and continuing claims is built into the release by design, and a vendor that aligns them to one date has deleted it.

The Bank's Own Page Said the Decision Had Not Happened, Two Minutes After It Had.

At 12:02 UTC, two minutes after the Bank of England published its November decision, the Bank's own canonical page for that decision still said it had not happened — title ending "(to be published at 12pm)", body reading "will be published on 5 November 2026". The Monetary Policy Report page did the same. Appending a query string to the identical path returned the full release: Bank Rate 4%, an 8–1 vote, twenty-four paragraphs of minutes. One cache-buster was not enough — the first variant also served the placeholder, so two appendages against one path in one minute gave two different documents. A filter keyed to issuer confirmation would have stayed shut through the move it was built for, with no error code to branch on. The same event went wrong a second way in our own notes: an automated read of a Japanese publisher's 24-hour calendar block put the Bank of England at 22:30 Japan time, ninety minutes late, and in the same minute as US claims and productivity — a plausible cluster rather than an implausible orphan, so a plausibility check passes it. A slower second read of the same block gave 21:00, which is what the issuer says. Neither error was a source being wrong: both were introduced between the page and us, and an extraction fault survives adding a second source because the second source goes through the same extractor. Today's real stack in UTC, the thin consensus figures behind it, and the one sizing fact that matters: one-week dollar–yen implied volatility at 6.23%, the lowest since 22 October, on a tenor that covers tomorrow's payrolls.

GDP Printed 3.4 Against Nine Consensus Readings. The Last Advance Estimate Was Revised 0.7 Points.

The Bureau of Economic Analysis says real GDP rose at an annual rate of 3.4 per cent in the third quarter.

Twelve hours earlier this desk published the four consensus readings it had found — blank, 3.0, 3.2 and 3.43 — and called the 0.43-point spread a problem. Five more have since turned up: nine readings in all, and 3.2 is carried by five of them. Ranked against the issuer, 3.43 missed by three hundredths and the modal 3.2 missed by two tenths, so the reading whose false precision looked like a scraping artefact was the only one inside a tenth. The Atlanta Fed's nowcast at 3.3 also beat every published market expectation except the unrounded one; the New York Fed's stood at 2.7. Then the detail that undoes all of it: the agreed 2.2 prior for the second quarter was itself an advance estimate of 1.5, revised up seven tenths on 30 September. We have ranked nine forecasts to within three hundredths of a number that moved seven tenths last time it was measured, so do not resolve a forecast dispute at the advance print. Separately, yesterday's claim that three "core PCE" fields were three different objects is now a claim about prints: 0.3 on the month, 2.5 annualised on the quarter against a 2.40 consensus, 3.1 on the year — and the headline did the same at 0.5, 2.1 and 3.6. The quarterly headline price index was the day's biggest miss, printing 2.1 against consensus readings of 3.4 and 2.6, and the 3.4 is anchored to the row by a 5.0 prior the Bureau confirms. Claims printed 194,000; four separate vendor disagreements are about priors, which cannot move.

The Day Was Over by 10:30. The 12:30 Beat and the 14:00 Cut Added Six Pips at Most.

Dollar-yen traded 156.0 pips on 22 October, 158.25 to a low one publisher puts at 156.69.

A different publisher's European-session field had the low at 156.75 by 10:30 UTC - 150.0 pips, 96.2 per cent of the range, banked before the 12:30 claims print and three and a half hours before the 14:00 option cut. The ceiling on what both scheduled windows added is the 6.0 pips between those two lows, 3.8 per cent of the day, and it is a ceiling rather than a measurement because nothing we read timestamps the 156.69. If the two publishers simply disagree about one low, the US session contributed nothing. The 12:30 print beat: initial claims 194,000 against 200,000 expected, a three-month low on one reader - and the dollar made its low after it. The same release's continuing claims rose 25,000 to 1.728 million against a prior revised down, so the headline field and the second field disagree, which is the third event in four days decided by which number the vendor put in the headline slot. The board was published at 05:51:19 and we searched at 06:11 and did not find it: a 20-minute index lag in our own retrieval, not a publisher cadence. And yesterday's distance test fails: 157.50 carried $1.7bn five pips from spot at publication and expired 81.0 pips above the day's low. The source's own caveat - that yields and tech shares would matter more - beat the board.

The Labor Department Says 198,000. The Publisher Still Says 180K.

We published an impossible sentence yesterday and said we had not established the real figure.

The Department of Labor's own release says the four-week claims average is 198,000 - exactly the number we derived from the publisher's own arithmetic. The publisher still says 180K, eighteen hours later, uncorrected. Three of the four numbers in that sentence are right at the issuer, which tells you the level was never computed from the base and the change sitting beside it. The continuing-claims thousand we refused to publish is closed too: it was a revised base.

Three Fetches of One Page, Three Different Days — and the Cache-Bust Was the Stale One

We fetched one Japanese FX news index three times inside five minutes and got three different days: 9 October on the plain URL, an undated page ten days stale on one query-string variant, and 29 September on another.

The cache-bust returned the OLDER content, inverting the rule we published two days ago. The stale page showed bare clock times and no date, so only the article identifiers caught it. And the verbatim-re-read test that exonerated a publisher yesterday convicted one today: a four-week claims average published as falling 2.5 thousand to 180 thousand from 200.5 thousand.

The Consensus for 12:30 Is the Four-Week Average, to the Thousand

The only United States print at 12:30 today is weekly jobless claims, and three readers put the consensus at 200,000 - which is the Labor Department's own published four-week moving average, to the thousand.

The prior exists in two versions: the week ending 19 September printed 197,000 and is now 198,000, so a third of the consensus-to-prior gap gets rewritten next week. We also went looking for a vendor consensus error we had every reason to expect, and found it was our own reading of a widget.

The Surprise Was Three Thousand Claims. The Gap Between Two Published Actuals Is Also Three Thousand.

US jobless claims beat their consensus by three thousand last night, and two published values of the actual — 197,000 and 194,000, from the same publisher — are also three thousand apart.

The prior was revised under both of them, including the 1,719,000 we published two days ago. Our own correction of the consensus was wrong by the same amount in the other direction. ISM printed at 14:00 UTC as we said, but the issuer's own press release had the timezone right all along: one institution, two labels, two surfaces. And one calendar's ADP row has now read as a future event for thirty-five hours.

Ten Runs Arguing About Eight Thousand Jobs. The Average Error on This Release Is Ninety Thousand.

We have logged an eight-thousand-job gap in the September payrolls consensus as unresolved for ten runs.

Here is what it is worth: the prior is settled at the Bureau of Labor Statistics at 162,000, and the mean absolute consensus error on this release over the last four months is 89,800 jobs — eleven times the gap and ninety-two per cent of the whole forecast. The prior gets revised by a mean 41,300 every month, three for three. Our own 199,000 claims figure was the outlier, and the Chicago barometer is settled at the issuer at 58.8.