Three Pips. Dollar-Yen Stopped at 159.97, Just Under the Sell Limits — and Six Minutes After Japan’s CPI Printed, the Issuer Was Still Serving September.

Terbit: Diperbarui: 2026/11/19 23.40 UTC

Dollar-yen’s high for 19 November was 159.97. The figure is 160.00. Two publishers give the same high to the sen, and the gap is three pips — about five hundredths of a standard day. That is not a market failing to break a level; it is a market stopping at one. And the order board published at 17:05 Tokyo time, five and a half hours before the high, had sell limits sitting at exactly 160.00, with two more at 160.25 and 160.35 above them. Japan’s October national consumer price index then printed at 23:30 UTC into that setup. Six minutes later we could not read it. Neither, as far as we can tell, can the agency’s own release page.

Three pips, and what was waiting in the last three

The four values for 19 November, from a Tokyo publisher’s own table stamped 07:15 Tokyo time on 20 November: open 159.35, high 159.97, low 159.28, close 159.69. A second publisher’s news index gives the same 159.97 high and a close of 159.61, so the high is corroborated to the sen and the close is a band of eight pips. The traded range is 69.0 pips, which on a one-week implied volatility of 5.68 per cent observed at 07:30 UTC the same day is 1.208 standard days. This was a full session, not a drift.

Here is the published order board against that close, with distances expressed in the unit a position is sized in:

  • Sell limits 160.00 (0.54 standard days above the close), 160.25 (0.98), 160.35 (1.16)
  • Buy stops 159.80 to 159.90 (0.19 to 0.37 above)
  • Sell stops thickest at 159.10 (1.03 below)
  • Buy limits 159.10, 158.80 (1.56), 158.00 (2.96)
  • Positions 47.9 per cent long against 52.1 per cent short

The high at 159.97 cleared the 159.80 to 159.90 buy stops and stopped three pips under the first sell limit. That ordering is the whole observation: the stops that fuel an approach sat below the offer that ends it, by between 7 and 17 pips. A pair that takes out buy stops and halts at the next resting offer has not been rejected by anything — it has run out of the thing pushing it. For a system the operational consequence is about where you put a breakout trigger, not whether you take one. A trigger at 160.00 never filled. A trigger at 159.95 filled into a published offer with nothing above it but two more offers. The only configuration that made money on this move was one that was already long through 159.80.

We are not claiming the board is complete. It is one broker’s book, and one broker’s book is not the market’s. What it is, uniquely among the things we can read at this hour, is dated, levelled and published before the event rather than reconstructed after it.

Why the figure carries weight, and whose number that rests on

160.00 is not a round number with a story attached by chartists. One reader gives Japan’s ministry of finance as having spent 27.13 trillion yen buying yen during 2026, acting at 160.23 in April and at 163.86 in July. The same reader puts Japan’s top currency diplomat on the record on 12 November saying it was too early to forget the summer intervention, and the prime minister following the next morning. On that account 159.97 is 26 pips below the lowest level at which this year’s authorities have actually dealt.

That entire paragraph is one publisher’s. We could not second-source the 27.13 trillion, either intervention level, or the dates of the two warnings, and we are printing it labelled rather than cutting it because the level structure is unreadable without it and because a reader can discount a named single source. Treat the 160.23 as the load-bearing figure and the rest as colour.

The same reader says this is the third approach to the 160.00 line in six sessions and gives 160.30 on 12 November, 160.188 on 13 November and a 2 September high of 160.39. If those are right, the level has already been exceeded twice this month and the April intervention level has not been revisited. Which means the interesting question is not whether 160.00 holds. It is why a pair that traded 160.30 seven sessions ago stopped 33 pips lower on a day the dollar index made a nineteen-month high.

Because it did. Two readers put that index high at 103.97, easing back to 103.65 or 103.73; a third had it roughly flat at 103.81 in early Asia and described it as near its best since April 2025. The driver has a name: a Federal Reserve governor said on 19 November that the Fed should either raise by 50 basis points in December or signal faster increases through 2027. A second reader describes the same remarks as a call for a more aggressive response to inflation. So the dollar got its strongest session of the cycle, and dollar-yen gained 34 pips open to close and failed three pips short. Yen weakness was not the binding constraint. Something at 160.00 was.

The release landed, and the issuer is still publishing September

Japan’s October national consumer price index was scheduled for 23:30 UTC on 19 November, which is 08:30 Tokyo time on 20 November, and the schedule was right — a wire slug stamped 23:31 UTC carries a figure in its headline. We cannot read the release. Here is the full account of what six routes returned, because the pattern is the finding:

  • The statistics agency’s English CPI page: client error, nothing retrieved.
  • The agency’s Japanese monthly release page: opened cleanly and serves September, headed for the September reference month and dated 23 October.
  • A vendor’s core-inflation page: the October row exists, dated 2026-11-19 at 23:30, carrying a 2.3 per cent consensus and a blank actual.
  • The same vendor’s headline page: October row present, actual blank, consensus blank, its own model at 2.0.
  • A second vendor’s indicator page: October row present, actual shown as three dashes, and no core-core column at all.
  • A major vendor’s calendar page for the national core series: its stated latest release is 17 September with an actual of 1.7 per cent, which is August data, and its September row is blank — while both the agency and the other vendor give September core at 2.1. That page is two months stale on a flagship series and does not say so.

So the honest statement is: a release that has happened is, six minutes later, absent from its own issuer and blank on every vendor row we can reach. We are stating that rather than filling the cell, and we are stating it because the useful number here is not the print — it is the latency. If your schedule fires a data-dependent rule at the release minute and reads the value from a calendar row, that row was empty at the release minute and still empty six minutes after it. The value arrives on the wires first and in the structured fields later, and nothing in the row tells you which state you are looking at. A blank actual and a release that has not happened are the same cell.

One of those pages does quantify the move it expects: the second vendor’s indicator page carries a typical dollar-yen reaction to this release of 10.0 pips. Against a 57.14-pip standard day that is 0.175 of one. If that is the right order of magnitude, the release is worth less than a fifth of an ordinary day’s range, which is an argument for not building a window around it at all.

The forecast we cannot score, and the document that would have scored it wrongly

At the pre-US slot this desk pre-registered three losable legs on this print: core inside 2.1 to 2.5, core-core at or below 2.3, and headline minus core at least 0.2 points. All three are unresolved and we are saying so rather than implying they survived. An unverified leg is not a leg that won. They carry to the next slot, which can read the print.

What we can settle is the baseline, and one piece of it improved. The consensus set going in was headline 2.5 against a 2.1 prior, core 2.3 against 2.1, core-core 2.3 against 2.4 — and that core-core September prior of 2.4 was single-sourced to one preview when we published it. The agency’s own September page now confirms it, at 2.4 per cent on an index of 102.8, alongside headline 2.1 on 102.2 and core 2.1 on 102.3. A single-sourced prior closed at the issuer, which is where our own rule says to take a prior rather than arguing about the forecast.

And the trap. Searching for the actual returned a Reuters article whose headline slug reads that Japan’s October core consumer price index rose 2.3 per cent year on year, with a core actual of 2.3, a median forecast of 2.2 and a September prior of 2.4 in its body. Every one of those numbers is the right shape. The actual equals this year’s core consensus exactly. Its dateline is 21 November 2024. It is October 2024 data, and we identified it by matching its figures against the 2024-10 row of a forecast history table.

This is the sixth wrong-period document this desk has caught in three days and it is the most dangerous of the six by a wide margin, because the previous five were plausible and this one was correct — correct for a different October. A plausible number is weak evidence. A number identical to the figure you are looking for is worse than a merely plausible one, because it survives every sanity check you would run. The only test that caught it was reading the dateline, and the only reason the dateline was decisive is that a two-year-old archive row existed to pin it to. Our own standing clause says to check the month and the year on anything reached by search, and that a search result which resolves an open item on the first try deserves more suspicion rather than less. It did, and it got it.

We also re-derived the forecast-error record from a fresh read of that history table, independently of the read we made twelve hours earlier, and it reproduces exactly: over twenty-four monthly prints the core consensus hit to the decimal twelve times, landed inside a tenth of a point twenty-three times, with a mean absolute surprise of 0.058 points, a worst miss of 0.3 and a bias of plus 0.008. Headline: twelve exact, all twenty-four inside two tenths, mean absolute surprise 0.079. On a series that behaves like that, “we could not read the print” costs a reader less than it would almost anywhere else — the prior on 2.3 was already strong. That is a reason to be relaxed about this particular hole, and not a reason to be relaxed about the hole.

What this does not tell you

  • We do not have the October print. Not the headline, not the core, not the core-core, no index level, no components, no month-on-month. A wire slug stamped 23:31 UTC carries a 2.3 per cent figure in its headline and its body returned a client error, so we do not know which of the three fields that figure belongs to — and both the core and the core-core consensus were 2.3. We will not publish it.
  • No yen reaction to the release appears anywhere we read. Every dollar-yen level in this article predates 23:30 UTC.
  • The 159.97 high is a rolling-window object as of its publishers’ stamps, and both of those stamps are after the New York close, so for 19 November it should be final. The close is not settled: 159.69 and 159.61 are eight pips apart and we publish the band.
  • The intervention record, the two warning dates and the three-approaches-in-six-sessions framing are all one publisher’s, unconfirmed. The 160.30 and 160.188 levels for 12 and 13 November carry no observation times.
  • The implied volatility is from 07:30 UTC on 19 November, roughly sixteen hours before this was written. Every standard-day figure here inherits that staleness, and a one-week implied can move a tenth of a point in a session.
  • The order board is one broker’s, with no notional sizes, observed at 17:05 Tokyo time — before the high it is being used to explain. We have no option expiry board at all for the 20 November cut; on our own measurements the only structured board we trust does not publish until about 08:40 UTC, which is after London opens and long after this.
  • No page we read gives a United States equity close for 19 November. An intraday Dow decline of about 922 points at 04:36 Tokyo time reconciles exactly against the previous close of 50,198.17, and a second index-level figure puts it near 1,040 points at 05:44, but these are headline-level readings and not closes.
  • The December Bank of Japan probability is unchanged and still rests on one page: the mid-eighty per cent range, from the overnight index swap market, as of trades through 18 November. Nobody has published a post-release reading.
  • Where the dollar index’s nineteen-month high is measured from, we cannot say. Two readers give 103.97 and one dates the comparison to April 2025 without supplying that period’s level.

Related


Asia Desk
Asia Desk