Last night this desk published that one page naming the CME FedWatch tool carried a 100 per cent probability of a December Federal Reserve hike at 13:00 UTC and 85 per cent at 16:30, three and a half hours later, across a print that made a hike more likely rather than less — and we called that a source contradicting itself. We also published that four readings of one named feed span 15.6 percentage points and that those numbers are therefore not a time series. Both claims are withdrawn. The two figures were never the same field. One is the probability of a hike of any size; the other is the probability of a quarter-point hike specifically. Our own hard rule says to ask whether two numbers describe the same object before treating a gap between them as a defect, and we did not run it on ourselves.
98.6 plus 1.4 is 100.0, and that is the whole explanation
A Tokyo news index carries an item stamped 12:54 UTC yesterday — thirty-six minutes before the consumer price release — giving the CME FedWatch probabilities as 98.6 per cent for a 25 basis point December hike and 1.4 per cent for 50 basis points. That is headline-level data: the publisher’s article bodies would not open for us this morning, so we label it as a figure read off an index and not out of a body. But the 1.4 is not single-sourced. We published it ourselves last night from a different publisher, feed named, stamped 13:00 UTC — two readers, six minutes apart, on the same decimal.
The two branches sum to 100.0. So does the pair we published on Monday: 84.4 per cent on a quarter point and 15.6 per cent on no change. Once you see that both pairs are exhaustive distributions, the “contradiction” dissolves. A page saying 100 per cent is reporting the chance of a move; a page saying 85 per cent is reporting the chance of a specific move. Nothing on either page is wrong.
And the reconciliation is not a hypothesis we are stretching to fit. Two independently published pairs landing on exactly 100.0 is not a coincidence, and the alternative reading requires one page to carry two values for one field three hours apart while a print pushes that field the other way. The field explanation requires no error anywhere.
The number that looked like it moved 15.6 points moved 0.6
Here is what the retraction costs us, stated as an arithmetic series rather than as a vendor complaint.
- Probability of a 25 basis point hike: 84.4 per cent on Monday around 11:40 UTC, 85 per cent yesterday at 16:30 after the print. A move of 0.6 points.
- Probability of no change: 15.6 per cent on Monday, absent from yesterday’s pre-print pair, which sums to 100.0 without it. A collapse of about 15.6 points — and it happened before the inflation data, not after.
- Probability of 50 basis points: 1.4 per cent pre-print, about 11 per cent in two Tokyo morning notes from one publisher. A factor of 7.9, from a base we had wrongly published as nothing at all.
So the 15.6 points we presented as a measurement failure is the distance between two different branches of one distribution, and the branch that actually repriced yesterday is the one we told you was empty. The useful and slightly uncomfortable version: the no-change branch died before the CPI, which means the print confirmed a repricing that had already happened rather than causing one. If you sized a news window around 13:30 UTC yesterday expecting the data to do the work, the data was late.
Three regional Federal Reserve presidents and a governor spoke during Tokyo hours overnight, between roughly 23:45 UTC on Tuesday and 00:30 UTC. One Tokyo index renders them as: more work to do, with no comment on the size of a move; a hike needed but no rush; and October’s consumer prices described as worrying. One of those we can read in a body rather than a headline — a Tokyo wire at 03:12 UTC has the Philadelphia president saying further hikes are needed to contain sticky inflation, on a gradual path that depends on conditions. None of them named a size.
The Bank of Japan makes the same field error about itself
We have carried an item for four slots: an issuer-sourced Bank of Japan expectation that core inflation runs above 3 per cent. This morning we read the issuer document. It is there, in the Summary of Opinions for the 29 and 30 October meeting, released at 23:50 UTC on 9 November under an embargo line naming 8:50 a.m. Japan Standard Time on 10 November. In the Prices subsection, one opinion projects the year-on-year rate of increase in the consumer price index, all items less fresh food, to exceed 3 per cent toward the end of the fiscal year.
Now set that beside the Bank’s own forecast table, decided at the same meeting and published eight days earlier. The Policy Board majority’s forecast for the same index in fiscal 2026 is a range of +2.4 to +2.6 per cent with a median of +2.5. The entire range tops out 0.4 points below 3, and the median sits 0.5 points below the opinion’s floor.
These do not contradict each other, and that is precisely the problem. A fiscal-year average and a terminal month are different objects: core can average 2.5 across the year and print above 3 in its final months. Both numbers are correct. But it means the figure the hawkish argument actually runs on does not appear in the forecast table at all, and anyone who has stored “the Bank of Japan expects 2.5 per cent” as the Bank’s inflation view has stored an average where the policy case is built on an endpoint. That is the same defect we just confessed to in the Fed probabilities, found twice in one morning, once in our arithmetic and once inside a central bank’s own document set.
The rest of that document, for the record: fifteen opinions in the monetary-policy section, no vote and no decision recorded, underlying inflation described as already at 2 per cent and expected to be anchored around it, and the neutral rate called likely to be considerably above 1 per cent with no range stated anywhere. The item is now discharged.
A footnote takes 1.3 points off fiscal 2027, and it is not in anybody’s forecast
The same forecast table carries a footnote that is worth more than most of the numbers above it. The Board’s projections exclude a proposed temporary cut in the consumption tax on food and beverages, and the Bank states its own estimate of that measure’s effect: about minus 1.3 percentage points on fiscal 2027 core inflation, and about minus 1.4 on core excluding energy.
Do the subtraction. The published fiscal 2027 median for core is +2.6; less 1.3 gives 1.3 per cent. The core-excluding-energy median is +2.7; less 1.4 gives 1.3 per cent as well. Both land in the same place, 0.7 points below the 2 per cent target, in the fiscal year the tightening path is supposed to run through.
We are not forecasting the tax measure and we have no view on whether it passes. The point is narrower and it is about what your system holds. A stored Bank of Japan inflation path of 2.5, 2.6, 2.0 is conditional on a named fiscal measure not happening, the issuer says so in a footnote, and the issuer has already published the size of the conditionality. If you are carrying a rate-path assumption keyed to those medians, it has a branch in it that the Bank has quantified for you and most calendars have not copied across.
Tonight’s producer prices: the whole expected move is inside the error bar
Japan’s Corporate Goods Price Index lands at 23:50 UTC tonight. The issuer’s September release states the next monthly report for Thursday 12 November and gives its release hour as 8:50 a.m. on the face of the document — without naming a zone anywhere. The Japanese indicator page we use agrees on 12 November at 08:50 and also states no zone, and the summer-and-winter conversion rule that page has previously carried in words was not on it this morning. So the conversion to 23:50 UTC rests on the convention rather than on either page saying it.
The consensus, from one source and labelled as such: +0.3 per cent month on month and +6.5 per cent year on year, against September’s +6.8 per cent and minus 0.3 per cent, both of which we have confirmed against the issuer’s own release.
Now the part a system trader should care about. We took the last six forecast-and-result pairs on that page and computed the mean absolute surprise ourselves: 0.40 points on the year-on-year field and 0.233 on the month-on-month. The expected year-on-year move tonight is 6.8 to 6.5, which is 0.3 points. The series’ own typical miss is 1.33 times larger than the entire change the consensus is asking for, and the expected monthly print of +0.3 is only slightly bigger than a typical monthly error. A blackout sized by how much the number is supposed to move would wave this release straight through.
One more thing about that page, which we publish as a limit and not as a finding: it carries a column for the revised prior, populated in six of its seven most recent rows, and in the September row that column reads minus 0.3 against a September result of minus 0.3. The page does not state the column’s convention, so we will not quote revision magnitudes off it. What survives is the shape: this series revises its prior inside the same release that carries the new print. A filter scoring tonight’s number against a stored September figure may find September moved in the same message. Japan’s October money supply, out this morning at +1.7 per cent year on year, did exactly that — its prior was revised to +1.9 from +2.0 in the same release. Two same-release revisions before the London open.
What this does not tell you
- We still have no dated numeric 200-day moving average for dollar-yen. Thirteenth consecutive slot. We did find something adjacent and better than nothing: a Japanese bank publishes a dated daily table of its published dollar-yen fixing for every business day of 2026, which is the first daily dated level series for this pair in our file. A 200-day average of fixings is not the 200-day average of closes everyone quotes, so we are not going to compute one and call it the 200-day. We are naming the series so the next slot can.
- The 98.6 is headline-level. It came off a news index whose article bodies returned errors on every attempt this morning. The 1.4 beside it is two-sourced; the 98.6 is not, and the reconciliation leans on it.
- The 50 basis point figure of about 11 per cent is one publisher’s two notes, half an hour apart. Two notes from one house are two desks, not two houses, and we do not count them as corroboration.
- The overnight Fed remarks are headline renderings from one Tokyo index, except the Philadelphia one. We have no timestamps on the remarks themselves, only on the items reporting them.
- We cut a central-bank claim entirely. One index headline attributes to the German central bank’s president the view that a December move is a fairly clear option. We could not open the body, and from a headline we cannot even tell which central bank’s December meeting is meant. So we are not publishing it as a fact, and we are not naming the official.
- No option expiry board. Seventh consecutive session. This morning we think we finally understand why, and it is a scheduling fact rather than a sourcing one — our sister piece this morning has it.
- The consensus for tonight is one source. We looked for a second and the best candidate a search returned was a preview of the November 2025 print, which is the date trap this desk writes about most often and nearly fell into again.
Where the level actually is
The one figure this morning that is dated, auditable and published by two institutions is the Tokyo fixing. A major Japanese bank’s published dollar-yen fixing for today is 158.87, and a second bank’s dated table carries exactly the same figure for 11 November, against 158.14 on 10 November and 158.34 on 9 November. That is 73.0 pips higher day over day on a series with dates on it, which is more than this desk has usually been able to say about a dollar-yen level.
158.87 sits 17.0 to 27.0 pips above the 158.60 to 158.70 band our undated 200-day readings span, and 10.0 pips above yesterday’s New York close of about 158.77. The Tokyo morning session, read in a body stamped 03:12 UTC, ran 158.66 to 158.95 — a 29.0-pip range whose top is 5.0 pips under the 159.00 handle that one Tokyo house says intervention wariness is capping. Euro-yen ran 177.18 to 177.48, euro-dollar 1.1158 to 1.1170, a 12.0-pip range. One house’s forecast range for the day is 158.000 to 159.400, 140 pips wide, with the fixing 62 per cent of the way up it.
We are not going to tell you which way it breaks. We will say that the branch which repriced yesterday was the tail, not the body, and that the release tonight is one whose consensus change is smaller than its own forecast error — so if you are holding yen risk into 23:50 UTC, the question is position size against an unforecastable print, not direction against a forecast.
