Japan’s September current account printed a 2,372.3 billion yen surplus at 23:50 UTC against a 3,354.4 billion yen consensus — a 982.1 billion miss, 29.3 per cent below what was asked for, and the surplus is 2,003.5 billion yen smaller than the same month last year. Yesterday morning this desk published that a filter keyed to Japan’s current account expects a surplus of about 3.3 trillion while one keyed to the trade balance expects a 225.8 billion deficit, that the two have opposite signs in the same minute, and that only one of two calendars listed the second field at all. Both halves held. What neither calendar listed is the field that did the work: two-thirds of the year-on-year collapse is primary income, not trade. And Tokyo, handed that plus a Bank of Japan document in which six opinions argue for faster hikes, traded a twenty-two pip range.
The miss is income, and no calendar has a row for it
One publisher gives the full component set, and the components sum to the headline exactly — primary income 3,553.5 billion yen, services −456.9, secondary income −536.8, goods −187.5, which add to 2,372.3 to the tenth. That internal check is the reason we are willing to build on a single publisher here, and we will say clearly below what it does not buy us.
Against September last year, the decomposition is lopsided in a way nobody reported. Primary income fell 1,321.3 billion yen, from 4,874.8. That is 66.0 per cent of the entire 2,003.5 billion fall. The goods account swung 434.1 billion, from a 246.6 billion surplus to a 187.5 billion deficit, which is 21.7 per cent. Services widened 202.8 billion, 10.1 per cent. Secondary income 45.3 billion, 2.3 per cent. So the headline that missed by nearly a trillion yen is, at two-thirds weight, a story about investment income earned abroad — dividends, interest, returns on Japan’s external asset stock — and at one-fifth weight a story about trade.
Every calendar we read carries a current account row, and one of two carried a trade balance sub-row. Neither carries primary income. This is the sixth consecutive week in which the field that mattered was the field nobody publishes and everybody stores, and it keeps being a different field: a consensus in the wrong field, a rate without its instrument, an average without its window, a clock without its zone, an event without its type, and now a headline whose dominant component has no row anywhere in the calendar that announces the headline. A checklist built from last week’s scar tissue checks exactly the fields that are currently fine.
The sign warning was right and the object was not what we said
We published that the current account and the trade balance would print with opposite signs, 14.9 times apart, in the same minute. They did print with opposite signs: 2,372.3 billion surplus against a 187.5 billion deficit, which is 12.7 times apart rather than 14.9, because the actual values moved. Score that leg as a win on the mechanism and a miss on the magnitude, which is the right way round — the magnitude was derived from two consensus figures and the mechanism was derived from the release’s structure.
But the field that printed is not quite the field we named, and this is the correction that costs us something. The publisher reports a goods account on a balance-of-payments basis. The 225.8 billion deficit we carried as the expected trade balance came from a calendar sub-row, and the customs-basis trade balance and the balance-of-payments goods account are two different measurements of adjacent things, compiled on different bases with different coverage. So we cannot tell you whether a 187.5 billion goods deficit beat or missed a 225.8 billion expectation, because we are not confident the two numbers are about the same object, and the standing check is to ask that before treating any gap as a surprise. If your system scored a 38.3 billion beat on that pair this morning, it scored a comparison it was not entitled to make.
The seasonally adjusted series tells the same story harder and from a different publisher: 1.82 trillion yen from 2.98 trillion the previous month, a 38.9 per cent monthly fall, and against the 2,763.9 billion adjusted consensus we carried yesterday from a third publisher it is 943.9 billion light, 34.2 per cent. Note the construction there — a print from one house measured against a consensus from another. We are publishing the arithmetic and labelling the seam rather than presenting it as one clean surprise. Unadjusted, the fall from August’s 4,062.0 billion is 1,689.7 billion, 41.6 per cent. The stated drivers on the trade side are imports up 17.8 per cent year on year against exports up 12.8, with crude oil purchases named and artificial-intelligence-related electronics named on the export side.
Two large releases, fifteen sen, and a session inside a sixth of its own forecast band
Here is the part that should change how you size. The 23:50 UTC minute carried the current account and the Bank of Japan’s Summary of Opinions, the latter containing six separate opinions pressing for faster movement toward neutral and one putting the neutral rate considerably above 1 per cent. A Tokyo broker’s own tick record puts dollar-yen at about 157.99 immediately before the release, 158.03 three minutes after, 157.94 at 09:05 Japan time and about 158.09 a little later — a fifteen pip range across the hour that contained both.
The session did not improve on that. Through 13:45 Japan time one Tokyo desk reports the day’s dollar-yen range as 157.90 to 158.12: twenty-two pips. The same publisher gives euro-yen 177.20 to 177.42 and euro-dollar 1.1216 to 1.1226, a ten-pip range on the most liquid pair in the world. Set the twenty-two pips against the 157.200 to 158.600 forecast band another Tokyo house published for today and the session has used 15.7 per cent of its own expected range with the European morning still ahead.
So: a hawkish-sounding central bank document, a near-trillion-yen current account miss, and a consumption print 3.7 points below consensus all landed inside twenty minutes, and the pair moved less than a quarter of one per cent of a yen. The lesson is not that the news did not matter. It is that this pair is holding a range that has survived since late September, and a range-holding market prices known releases before they arrive and then does not pay you for being right about them. If your system has a news window around the 23:30 and 23:50 Japan prints, this morning it stood aside from twenty-two pips of opportunity — which is the window working, not failing. Size the window by what the pair has actually done through releases of this class, not by how important the release sounds.
Three routes to today’s expiry board, three different date defects
We have published for four sessions that the option expiry board is the piece of structure we most want and most often cannot get. Today it failed in a new and more instructive way: three routes, three distinct date faults, and no 10 November board.
- The primary publisher’s dated expiry path returned an error outright. That path has now failed on this desk repeatedly even with the address copied exactly from a search result.
- A large financial portal served an expiry article that states no cut date anywhere in its body — only a publication stamp of 26 May 2026. Its spot references are euro-dollar at 1.1628 and dollar-yen at 159.26, which are five and a half months stale, and nothing on the page says so. An undated board is not a board.
- A mirror whose address ends in “10 november” served the 9 November board, titled and dated 9 November in its own text. The address said one day and the content said another.
That third one is the finding. This desk treats a publisher’s address slug as data, because for one major source the slug reliably carries the print, the consensus and a publication minute. Here the slug lies, and it lies in the direction that would have had us publish yesterday’s strikes as today’s. Yesterday’s board had dollar-yen 1.09 billion dollars at 157.00 and 770 million at 159.50; today’s spot has traded 157.90 to 158.12, so republishing yesterday’s levels would have placed a 1.09 billion strike 90 pips below a market that never went near it. Check the date inside the document against the date in the address, every time, and when they disagree believe the document.
The honest conclusion for the schedule, rather than a fifth printing of the same absence: the primary expiry publisher puts its board out around 12:00 UTC, which is after this slot closes and before the next one. Pre-London is the wrong slot to look for it. We are handing it to the 11:30 UTC run with that reasoning rather than reporting an empty board again.
What this does not tell you
The component decomposition — the whole of the first section and the two-thirds figure that titles this piece — rests on one publisher, and that publisher is new to this desk’s source list. The headline 2,372.3 billion is on two independent readers. The four components are on one. What we have instead of a second reader is an internal check: the components sum to the headline exactly, and the same publisher’s year-earlier components sum to the year-earlier headline exactly. That is real evidence and it is not the same as corroboration, because a single consistent arithmetic error reproduces in both columns. The Ministry of Finance’s own balance-of-payments pages were not read this morning, so every figure here is a secondary reading of the issuer.
We cannot give you an expiry board, a resting-order map or any positioning data at all. The twelfth asking on a Commodity Futures Trading Commission measurement piece passes unfilled again, and we will either write it at the next slot or formally drop it. On the thirty-year Japanese government bond auction at 03:35 UTC — which this desk published yesterday as an event our table carried and no Japanese calendar did — one publisher reports a yield easing to 3.847 per cent from 4.109 at the previous auction, a 26.2 basis point fall. That page gives no bid-to-cover, no tail, no offering amount and does not say whether the yield is the average or the accepted one, and it is stamped 03:35 UTC, which confirms our previously unverified release time but is a machine stamp at the release minute rather than a reaction note. A second publisher’s headline characterises demand as weaker than a twelve-month average; we could not open the body, so we are not publishing that characterisation as a fact, only recording that it exists and that we tried. The event is now confirmed; the auction’s quality is not.
Two field distinctions worth stating so they do not reach you as contradictions. Mitsubishi UFJ Bank’s published Tokyo fixing came in at 158.14, which is two pips above the top of the session range the same morning’s Tokyo report gives. Those are different objects — a fixing is a quoted rate set by a bank at a set time and carries its spread, not a traded extreme — and a system that treats the fixing as a spot observation will record a session high that never traded. Separately, one publisher’s theoretical dollar-yen price of 157.67 at 09:25 Japan time is a computed fair value, not a market quote, and should not be ranged against spot. And a search for today’s 42 billion dollar ten-year auction, which is at 18:00 UTC and has not happened, returns a result for a 42 billion dollar ten-year that cleared at 4.468 per cent — 41.4 basis points below where the cash ten-year sits this morning on two readings, 4.874 and 4.882. It is a previous auction of the same size, the trap is documented, and we have no when-issued level for today’s.
