Yesterday morning one publisher called 157.00 a level that could pin dollar-yen through the European session, with spot under fifteen pips away. This desk published that as a falsifiable test and named tonight as where it settles. It failed in both directions: the pair traded 36 pips below the strike and closed 38 pips above it. We also, for the first time on this channel, have the notional sitting on that strike — 1.4 billion dollars — from a source we had not been reading. Tokyo opens 157.34, and the Tankan has just printed at plus 24 against a plus 25 consensus — a two-point improvement that is still a miss.
The pin failed, and the notional was obtainable all along.
Minkabu publishes the New York option expiry ladder with sizes attached, at around 20:40 Japan time each day. Last night’s carried USD/JPY 157.00 at 1.4 billion dollars, 158.00 at 918 million and 159.30 at 1.3 billion, alongside a much heavier euro-dollar ladder topped by 2.0 billion euros at 1.1450 and 1.7 billion at 1.1300.
This matters beyond one strike. This desk has written five times that option expiry notionals are locked inside an image on the page we habitually read and cannot be extracted. That was true of that page and false as a general claim. A second publisher has been printing the same ladder in plain text with the sizes on it, and we were not looking. We are recording that as our error, not as a discovery.
As for the pin: dollar-yen fell to 156.64 during the New York session and finished at 157.38. That is 36 pips through the strike on one side and 38 pips through it on the other, on a day when the cut was supposed to hold it. A 1.4 billion dollar strike is a real size and it did not organise the session.
We cannot settle the ten-minute window, and we looked.
The test as this desk wrote it was narrower than the day range: did dollar-yen print inside fifteen pips of 157.00 in the ten minutes before the 14:00 UTC cut. We went looking for that and no publisher we read gives a ten-minute window. The session summaries give opens, highs, lows and closes. The 156.64 low and the 157.38 close both sit outside the band, and the path between them crossed the strike, but where price actually was at 13:55 is not a fact we can put in front of you.
So the honest verdict is split. The pinning claim, read as “the cut will hold price near 157.00 through the European session”, is refuted on the day range. The precise test we set ourselves is unsettleable from public sources, and setting a test that requires tick data you do not have is a mistake in test design, which is ours.
The move was American and it reversed inside two hours.
The August core PCE deflator printed soft at 12:30 UTC — plus 0.2 per cent on the month and plus 3.0 per cent on the year against plus 0.3 and plus 3.3 expected. October rate-hike expectations were pared, long yields fell, and the dollar was sold. Then the Chicago activity print landed strong and crude rebounded, yields turned back up through 5.30 per cent on the ten-year, and the dollar was bought back. Net of all of it, dollar-yen opens Tokyo at 157.34 to 157.35, 0.44 yen lower on the day.
Two things for a system trader. First, the round trip spanned roughly two hours and the day ended near the top of the range, so a rule that took the post-PCE dollar sale as a signal was on the wrong side by the close. Second, the two readers of the session high do not agree: Fisco has 157.38, Gaitame has 157.53. Fifteen pips. They agree on the low to the pip. If your stop or your range filter is anchored to a published session high, it is anchored to a publisher.
On the rate path itself, the numbers being quoted for October are 34.9 per cent for a hike on the CME FedWatch reading carried by Fox Business after the print, and 37 per cent at Gaitame. This desk published 47.1 per cent yesterday morning, citing the same FedWatch tool at one remove. A ten-to-twelve point repricing on a soft core print is credible. We still have not read that feed directly, and the fact that we are quoting it at second hand for the tenth consecutive slot is a standing embarrassment.
The Tankan printed plus 24, and two releases arrived in one minute.
At 23:50 UTC the Bank of Japan published the September Tankan and the Summary of Opinions from the September policy meeting simultaneously. The large manufacturers’ business conditions index printed plus 24, against a consensus of plus 25 and plus 22 in the June survey. That is a two-point improvement and a one-point miss at the same time, and which of those two facts a headline leads with will depend on the headline. This figure is a single calendar read taken three minutes after the print and it is one source; a second publisher’s page had not populated the row at all when we looked. The rest of the preview consensus, which we have not yet marked to market, was plus 36 for large non-manufacturers against plus 37, and capital spending plans at plus 12.2 per cent against plus 11.5.
The scheduling is the trade-relevant part. Two separate Japanese policy inputs arrive in the same minute, ten minutes before the Tokyo fixing hours get going, in a book that has just absorbed a full American session and has not yet handed to anyone. If you run a news filter on Japanese releases, this is one window, not two, and it needs to be wide enough to cover a Tankan surprise and a hawkish or dovish set of board opinions arriving together. Splitting them into two five-minute blocks will leave you live in the middle of a single event.
157.35 against a corporate planning rate of 152.57.
The item in the Tankan this desk will be reading first is not the headline index. It is the assumed exchange rate. Japanese large firms built their current fiscal-year plans on 152.57 to the dollar. Spot opens this morning at 157.34.
That is 4.77 yen of gap, and it points in the direction that flatters exporter earnings and worsens the import bill, which is the same tension the Bank has been talking around since the September hike. Whether the assumption moves in this survey tells you something the business-conditions index does not: whether the corporate sector has accepted this level as the new base or is still treating it as an overshoot. The same release also carries firms’ own price expectations, which in June ran at plus 3.7 per cent for their own selling prices one year out, the highest since the survey began asking in 2014. Those lines are what a rate decision on 29 and 30 October gets argued from.
Thirteen hours on, the Ministry’s English page still has not said zero.
One follow-up from yesterday, because we said we would check it. Our Macro desk published at midday that the Ministry of Finance had reported zero intervention for 27 August to 28 September, that three wires carried it inside thirteen minutes of 10:01 UTC, and that the Ministry’s own English page did not have it. At 23:12 UTC we looked again. The English monthly index still lists 30 July to 26 August as its most recent entry, and the URL the Ministry’s own naming pattern predicts for a 30 September release still returns a 404.
That is thirteen hours and eleven minutes. The observation is now considerably stronger than it was at midday, and the rule it supports is worth repeating: use the issuing agency to settle a number, never to learn one.
What this does not tell you
We do not know where price was at 13:55 UTC and we have said so. We do not know which session high is right. We have not read the CME feed directly and we are quoting two second-hand values for the October probability that are themselves two points apart.
The Tankan consensus figures above are forecasts collected by one broker’s preview and they are not this desk’s own survey. The 152.57 assumption is the June reading; if it has been revised in this survey, the gap we describe is wrong by however much it moved, and that will be visible within minutes of this being published.
Nothing here is a view on direction. A pin that failed tells you the cut did not organise the session; it does not tell you the next one will not. And an opening quote is not a forecast.
