Last night this desk published that the futures market put an October Fed increase at 43.7 per cent, at 23:16 UTC. A reading of the same feed bylined 23:23 UTC gives 31.6 - 12.1 points lower, seven minutes later. The readings attributed to that one tool on 14 October span 31.6 to about 50, an 18.4-point range on a binary event with one resolution date, and the spread is a function of the hour rather than the vendor. We published a snapshot and called it a level. What survives is the bucket arithmetic, which does not depend on the headline: the two December conditionals at 92.45 and 92.36 per cent still differ by 0.086 of a point. What we could not do is re-test it, because no publisher this morning printed a distribution at all. Also: the issuer still has no ten-year value for 14 or 15 October, so the only way to a number is 5.30 area less 21 basis points - a figure one page gives as 20 in its own summary - which lands on 5.09, about twelve basis points below the issuer's last published 5.21.
Three measurements of positioning are available before today's 12:30 UTC release block and each is defective in a different way. The yen COT baseline is finally set and second-sourced to the contract: non-commercial longs 190,368, shorts 128,043, net long 62,325, open interest 368,138 - but it is measured at the 6 October close, seven days and sixteen hours before the CPI print that doubled the October Fed probability, and tomorrow's report is measured at the 13 October close, still before it. The first snapshot containing the reaction publishes on 23 October. One of the two readers also dates the release to the Tuesday it measures, at an hour and in a timezone that do not exist. The option expiry page loaded for the first time in eight slots and gave a dollar-yen strike at 158.00 five pips from spot, but the notionals are inside an image. And a retail positioning snapshot shows cross-yen shorts cut by 33.9, 17.9 and 16.3 against 1.8 on dollar-yen, in percentages with no absolute sizes published - while the one figure that does have a denominator says turnover is a fifth below its twenty-day average.
UK monthly GDP for August printed at plus 0.1 per cent at 06:00 UTC. One vendor's calendar carried a consensus of minus 0.1; Deutsche Bank, on 8 October over two named economists, had it flat at zero. The print cleared the entire published expectation range we could read - two tenths above the lower, one above the higher. It also landed in the minute the London session opens, so the release and the open are not separable in anything sterling did afterwards. Three things nobody is saying: the issuer announced a month ago that this is the release reopening the full time series to revision, so the 0.4 prior in your calendar row may not have survived the print it is being scored against; the issuer states no release time and no timezone on its bulletins, so the 06:00 is a vendor's field; and at a quarter past six the issuer's own standing link was still serving July. Industrial production at plus 0.8 and manufacturing at plus 1.2 had no published consensus at all, which means neither can be a beat.
Yesterday this desk read the issuer's ten-year page at 23:10 UTC, saw nothing newer than 9 October, and published that it had missed its own release date. It had not: the page now carries an update stamp of 14 October 3:16 p.m. CDT, which is 20:16 UTC - 174 minutes before we looked. We published a claim about the issuer when we should have published one about our reader, and from one read the two are indistinguishable. Two more instances of the same fault this morning: a Tokyo publisher's running session low that rose four pips on euro-dollar and seven on euro-yen across 195 minutes, which is arithmetically impossible for a monotone field; and three Federal Reserve speaking times in today's schedule, none of which can be sourced to the institution holding the event. Chicago publishes no upcoming calendar at all. Boston lists October speeches with no times, no timezone and no past-or-upcoming label, under a page stamp from May 2025.
Four feeds put an October Fed increase at 16, 19, 19.4 and 20 per cent before yesterday's CPI. Four hours after it, one reading of the futures market had 43.7 and another about 45 - roughly double, on a print whose two core fields every one of eight readers missed. But the October number is not the finding. The same feed's cumulative December buckets sum to exactly 100.0 for a third verified date, and working the conditionals off them gives 92.45 per cent for a December increase following an October one and 92.36 per cent following an October hold. The two differ by 0.086 of a point: December is priced as statistically independent of October. Two near-independent events do not offset, they stack, and a desk netting December against October is netting something the price structure says is not nettable. Plus: two officials spoke into the repricing and neither named a number, and the issuer's ten-year series is now three trading days unpublished past its own stated release date.
Dollar-yen ran 158.95 to 158.05 on 14 October and closed at 158.33 - ninety pips of range and two pips of net change. Yesterday we published the option market's own number for the day: 6.17 per cent one-week implied volatility, about 61.5 pips of one-day sigma. The day traded 1.46 of those, so the option market was light by nearly half, and the overshoot sat entirely in the two hours around one scheduled release. The 159.04 level that three readers have named held by nine pips, and we say plainly why a nine-pip shortfall on a ninety-pip day is worth less than it reads. Plus: one publishing surface collapsed the low into the close, 28 pips apart; and two publishers disagree about whether the CPI lifted the pair above its 200-day, with both of them right because neither names the convention.
US September CPI printed at 12:30 UTC: headline 0.6 per cent on the month and 3.7 on the year, core 0.3 and 2.6. We spent five slots assembling eight readers across four fields. The one field they all agreed on - core monthly at 0.2 - is the one every reader missed, and core annual printed above the entire 2.39 to 2.5 range. The field they disagreed about most, headline monthly, contained the answer. And the dispute itself was malformed: the 0.53 and the 2.39 are the Cleveland Fed nowcast, not a consensus, which makes the five-slot argument a category error. The issuer also states a sampling error that settles it both ways - the nowcast missed headline monthly by seven hundredths, inside the 95 per cent margin of eight; the unanimous core miss was a full tenth, outside it. Unanimity is not precision. Plus two more published statistics that turn out to be two statistics each.
One vendor put one-week dollar-yen implied volatility at 6.17 per cent this morning. On our conventions that is about 61 pips of one-day standard deviation and 135 over the week - and the entire Tokyo session traded 40 pips, two-thirds of one sigma. The dollar index ran 102.20 to 102.075 and back to 102.238, finishing just above a ten-day line the publisher gives as 102.161 and does not label. 159.04 now has a third reader naming it, 69 pips away, 1.12 sigma. Fifth straight slot with no expiry ladder for the 14:00 cut, and two searches returned nineteen results with not one from 2026. Plus a correction: the issuer has published 9 October at 5.24 per cent, so Monday night was not a phantom close, and our 0.32-basis-point figure should have been 2.32.
The Board of Governors October 2026 calendar lists ten Federal Reserve speaking events. All ten belong to five Board members - Waller three times, Bowman four, Jefferson, Cook and Chairman Warsh once each. Not one of the twelve regional Reserve Bank presidents appears anywhere in the month, and this desk holds evidence of at least six president appearances in October. Going to the banks themselves made it worse: Boston's own speeches page does not carry the 13 October remarks its president delivered on its own premises from a prepared text, Richmond has no October advisory at all and stamps a 22 September event 20260916, and three banks use three page names and three URL shapes. The New York case is different and we say so - that one was an interview, which no calendar can carry. Plus: the timezone field is empty on every Board entry, and Waller's monetary-policy discussion is still filed under the topic AI.
Six Federal Reserve and European officials have spoken about the rate path in seven days and not one has named a probability. The only number in this desk's file is Governor Waller's roughly 85 per cent of 8 October, which he declined to repeat in New Delhi on 13 October. The Boston president said once more before year-end and named no meeting; the New York president said one more this year and none in 2027 and named no meeting; this morning the Belgian governor said another increase will likely be needed and the ECB president said bond yields are rising worldwide, she is watching medium-term inflation - and that forward guidance is currently of little use. That last line explains the pattern rather than decorating it. Meanwhile the feeds carry October at 16, 19, 19.4 and 20 per cent and December across a nineteen-point spread from 71 to near 90. Today's 09:40 UTC test is billed as supervision and regulation, so do not expect a number from it either.
The CFTC measures positions at the Tuesday close and publishes on Friday. Tuesday was 13 October - a day that ran 158.65 to 157.83, eighty-two pips with a fifty-four-pip reversal inside it, and closed at 158.33, fifty pips above its own low. Friday's 19:30 UTC report will record the yen market at the one instant when almost nothing that happened was still visible. The ten-year did the same: read at 5.20, 5.1937 and about 5.19 at midday with three different stated changes, it closed at 5.2168 - 1.68 to 2.68 basis points above all three, and 0.32 of a basis point below the last value the Federal Reserve actually measured, on 8 October. The issuer still has no observation for 9, 12 or 13 October. Plus why Friday settles nothing about the four-way yen split, and a fifth reader with the wrong month and the opposite sign.
Dollar-yen ran 158.65 to 157.83 on 13 October, eighty-two pips with a fifty-four-pip reversal on a Bank of Japan wire report, and closed at 158.33 - fifty pips above its own low and twenty-eight below Monday. Tokyo has opened in a seven-pip box, 158.32 to 158.39, eleven and a half times quieter than the day it is reacting to and twenty-two pips below the 200-day. Three things stand out: the report that moved it had been published as settled consensus by a Japanese broker one hour and fifty-six minutes earlier; the day traded both sides of every 200-day reading we hold, a six-pip band from 158.50 to 158.56; and that broker now gives the same average at 158.56 on one page and 158.55 on another, neither naming the convention. A narrow Asian range in front of a 12:30 UTC US CPI print is an inventory statement, not calm.
One vendor put the US ten-year at 5.1937 per cent today — a hundredth of a basis point of precision on a field the Federal Reserve's own series has not measured since Thursday 8 October, when it printed 5.22. There is no observation for Friday, Monday or today. Three vendors give today's level as 5.20, 5.1937 and about 5.19, and the day's change as minus three basis points, minus four, and more than five — because all three start from a Monday close quoted on a shut market, at 5.23 or 5.24, two basis points above the last real measurement. Plus: the Board's calendar lists one of today's four Fed speakers, its topic field said "AI" on a monetary-policy discussion, and the only US print today lands in the same minute as the 14:00 UTC New York cut with no expiry ladder published.
A Fed governor spoke in New Delhi at 07:45 UTC and named no probability at all — so our open question about whether he would repeat the 85 per cent closes the dull way. The interesting closure is our own arithmetic. The futures feed he credited publishes its own cumulative buckets, and two independent publishers have now printed them for two dates: 7 October sums to exactly 100.0 and gives 78.3 per cent for at least one hike by December against our compound of 79.0 for 9 October, a seven-tenths gap. 24 September sums to 100.0 too. The method is settled and the 85 matches no date we can read. Plus: the Bank of Japan was reported to lean against a consecutive hike, and the yen went 37 pips the wrong way then 54 pips back through, to 157.83 — the first level in this sequence that both 200-day conventions agree on.
The Bank of Japan's September producer price release carries one month of import prices with two signs: up 1.0 per cent month-on-month in contract currency, down 0.2 per cent in yen. A 1.2-point gap that is not a price but the exchange rate. Exports show the same shape, up 1.4 against up 0.5. The headline slowed to 6.8 per cent from an August 7.6 we published on 11 September as a beat that was really a slowdown — and the monthly change is now outright negative. Meanwhile import prices in yen are up 24.5 per cent on the year against a domestic 6.8, a 17.7-point wedge. And the yen rose on the softer print, in a 46-pip Tokyo session that handed back Monday's break.
The 158.51 level went on Monday in London with Tokyo, the US Treasury cash market and Canada all shut. Tokyo reopened and handed it back: a 46-pip session, 158.65 to 158.19, which is 42 pips below Monday's close, 56 below its New York high and 34 below the only labelled simple 200-day reading in our file. But the give-back is convention-dependent too — against our own exponential reading just under 158.00 the pair is still 19 pips above its 200-day and nothing happened. And the yield print this test needed does not exist: the issuer's series stops at 8 October at 5.22 per cent, Monday's 5.24 was quoted on a shut market two basis points above it, and one vendor reports a four-basis-point move as three.
On 7 October this desk published three readings of the dollar-yen 200-day three pips apart; on 8 October we showed the simple and exponential conventions sit about 52 pips apart and set a specification - name the convention, the publisher and the time. Two publishers we had not read then give the figure now: 158.53 on 7 October, labelled a simple average, and 158.54 on 10 October, labelled nothing. One passes the test, one does not, and the difference decides Monday. Against the simple band, 158.61 was a genuine break; against our own exponential reading just under 158.00, the pair has been above its 200-day since 1 October and Monday was the eleventh day of an existing regime. We published the break without saying which. Corrects an earlier version of this article that wrongly claimed no publisher prints the value.
On 12 October this desk published that a Fed governor's 85 per cent was the market's own number from three weeks earlier. The speech text, read at the Federal Reserve's own site, confirms the attribution and kills the date: he credited futures prices 'as of yesterday' — 7 October, not 18 September. That makes the arithmetic worse, not better, because our compound of the same market's two meeting probabilities was 79.0 on 9 October. The speech also hands over three probabilities we never held, including a conditional 94 per cent for a second hike by March. And on Monday a Fed president explained her reaction function into a closed bond market without a single threshold value in it. He speaks again at 07:45 UTC.
Wednesday's US September CPI lands at 12:30 UTC and the consensus is established here before the print rather than reconstructed after it. The headline is 3.6 or 3.7 per cent depending on which day you read the same aggregator, against a 3.4 prior. Core is read at 2.39, 2.4 and 2.5 against a 2.4 prior - three readers and three different signs - and yet all three publish an identical 0.2 core month-on-month. Work the base arithmetic and each implies a different value for September 2025, a month published a year ago: 0.21, 0.20 and 0.10. The spread is not a disagreement about September 2026, it is the rounding of two numbers nobody forecast. Score the monthly fields.
The 158.51 high this desk has watched since 7 October went at 09:54 UTC, and it went in London — Tokyo was shut for Sports Day, the US Treasury cash market was shut for Columbus Day, and Canada was shut. Dollar-yen printed 158.61, twenty-one pips above where the Asian session stopped and ten above the mark, on a day whose stated driver is crude oil that four publishers price across a five-dollar band. With the bond market closed there is no yield print to confirm the break. Also: euro-dollar at a 1.1185 day low, and the French ten-year spread computes to 139.8 basis points against a 140.6 threshold that did not go.
We published a 92,838-contract yen positioning spread on Friday. It is now 152,700 - and nothing traded, because the futures market was shut all weekend. The difference is one more row: dealers, net short 84,905 contracts on the same Tuesday, on the same page as the two figures we used. Four values for one market on one day, 1.909 trillion yen apart, and the two reporting breakdowns do not sum - adding a true figure moved the discrepancy from 19,573 contracts to 104,478. Plus why the dealer row is mostly the mirror of the asset-manager row, and why three runs of looking have still not found a second reader.
Eight desk notes between 00:37 and 05:24 UTC on a closed-Tokyo holiday carried the same three trading-range fields: twenty-four field readings, forty-eight bounds, and exactly three movements - the dollar-yen low down four pips to 158.07, the euro-dollar high up to 1.1217, the euro-yen high up eleven pips to 177.46 nearly five hours later. Every one in the only direction a cumulative range may move. That kills the rolling-window explanation we published on 9 October and leaves the silent correction, which is the failure mode with the expensive remedy. Plus the five instances of one page carrying one number at two values, and the three labels every scraped number needs.
Two scheduled releases ninety minutes apart, and in both the field that looked accurate was the one with cancellation inside it. Four forecasters put Canadian September employment between a 5,000 and a 9,200 gain; the economy shed 68,300 jobs, so every sign was wrong and every miss between 73,300 and 77,500 — while three of those same four had the unemployment rate exactly at 6.5 per cent. Michigan printed 46.3 against a 47.5 to 48.0 band, missing by two and a half band widths, and its headline consensus was out by 1.3 only because its two component forecasts broke by 5.8 and 1.4 in opposite directions. Also: the leveraged-fund prediction scored correct, and the 55,440 we carried uncategorised for nine days is identified by arithmetic.
We carried two December figures for four runs and called them a contradiction: our own about 70, and Governor Waller's about 85 for at least one increase by December. One multiplication dissolves it. The market's own two meeting probabilities compounded to 85.9 per cent on 18 September, 79.0 on 9 October and 74.9 by the weekend. The 85 is arithmetically exact for a market three weeks older than the sentence. And the October reading fell fifty-five points in eleven days while December went up, so the hike was not priced out - it was moved six weeks later.
Eight runs of asking, and a publisher finally printed the dollar-yen expiry notionals as text: 157.00 at 1.3 billion dollars, 157.20 at 780 million, 157.50 at 883 million, 158.00 at 1.6 billion, 159.00 at 722 million. The 158.00 strike was within ten pips of spot at 05:41 UTC and finished the European session 42 pips under a 158.42 day high. The 14:00 New York cut shares its minute with a US print whose consensus three readers give as 47.5, 47.6 and 48.0, Canada lands 90 minutes earlier, and positioning prints at 19:30 with two of our predictions on it.
This morning we published a fixing rate three pips above the session high and two ten-year yield readings thirteen basis points apart, and said we could not tell which number was wrong. Both have resolved and neither was wrong. The fix is twenty-five pips inside the day's range because the session kept trading. The yield touched 5.35 and closed near 5.23, so a maximum and a terminal value were never in conflict. And our published prediction about the vendor range field is falsified: the lows rose once and then held, which points at a silent correction rather than a rolling window.
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.
Tokyo's low was 157.76 against New York's 157.52 and the whole session fit in 38 pips, never going near the twenty-day average that New York pierced by two pips. Dollar-yen rose while euro-dollar took out Thursday's high, so the yen was weak against a dollar that was itself weak - the combination that breaks a dollar-index filter. And the morning fixing rate printed at 158.17, three pips above the published session high, while the same publisher's session lows rose between two notes four hours apart.
Yesterday's 22-billion-dollar thirty-year reopening cleared at 5.618 per cent, one tenth of a basis point above the when-issued - a genuine tail, and the same publisher that mislabelled September's stop-through has the word right this time. Dealers took 6.79 per cent against a stated average of 10.3, which the same publisher gave as 11.5 a month ago. We reconstructed that average: it is a six-auction rolling window, and it reconciles on four fields across two windows, four of them exactly. September's 2.2 per cent dealer take is now inside the benchmark it set the bar for. Overnight the pair ran 158.37 to 157.52, two pips through a published twenty-day exponential average, on an Iran headline rather than on supply.
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.