Yesterday morning this desk published a piece headlined that USD/JPY had closed lower a fifth straight day into a wall of dollar-positive news, and asked what Tokyo does with a tape that ignores its own fundamentals. Day six answered it. USD/JPY traded 153.85 to 154.67 through New York, the dollar index pushed back above 99.00, and Gaitame.com’s morning note puts today’s expected range at 153.600–155.300 with 155 as the level being tested. The wall did not get taller. The market simply stopped standing on the other side of it. And later this hour, at 23:50 UTC, Tokyo prints its own producer-price number — expected at 7.4% year on year, two full points above the American one that did the damage.
Nothing new was added to the dollar case. It just got paid this time.
Here is what was on the list yesterday and is still on it today: US producer prices for August printed +0.4% on the month and +5.4% on the year, against a consensus we published at +5.3%. The Bureau of Labor Statistics attributes over three-quarters of the rise to final-demand energy, which moved up 4.2%, with diesel alone up 24.1% and responsible for more than a third of the goods index gain. Crude went with it: Brent is quoted between $105.44 (Investing.com, +4.18%, session range $100.21–$105.84) and $108.87 (Trading Economics, +6.34%), WTI between $100.50 (OilPrice, 15:50 UTC) and $103.76 (Trading Economics, +6.69%). The 10-year Treasury yield is 4.91% on Yahoo Finance’s close and 4.97% on Trading Economics, which calls it the highest since October 2023; the 30-year is 5.35% to 5.37%.
Every one of those inputs existed on Tuesday and Wednesday in a weaker form, and USD/JPY fell anyway. Yesterday we said that a market refusing free money is telling you something about positioning rather than about rates. That reading has now been tested and it held: what changed overnight was not the fundamental case but the willingness of a crowded short-dollar book to keep sitting in front of it. Six sessions of one-way yen buying, then a squeeze on the sixth. The information was in the position, not in the print.
Note the one thing that was genuinely new. US initial jobless claims came in at 206K against a 206K consensus — exactly in line, continuing claims 1.774 million. The labour market did not move the dollar. Energy did.
And the audited oil number, which we spent two articles arguing about the timing of, finally landed at 16:00 UTC in its holiday-shifted slot. US crude inventories fell 0.4 million barrels for the week ending 4 September, to 424.1 million — bang on the five-year average — against an API estimate of −0.3 million the night before. Gasoline built 1.3 million, distillates 2.1 million, and refineries ran at 97.8% of capacity on 17.6 million barrels a day of crude inputs. A draw that small, on those run rates, is a demand-neutral number. It moved nothing. Brent still added four to six percent on the day, which tells you the price is being set by the supply headline and not by the one inventory series anybody actually audits. That was the argument our Flow desk published yesterday morning, and the tape spent the afternoon agreeing with it.
Tokyo’s own producer prices are running two points hotter, and the yen is being sold anyway
Japan’s August corporate goods price index lands at 23:50 UTC — 08:50 JST, inside this session. Minkabu’s indicator page has consensus at 0.0% month on month and +7.4% year on year, against a July actual of +0.1% and +7.2%. The series has gone 2.6% in March, 4.9% in April, 6.3% in May, 7.1% in June, 7.2% in July.
Put the two side by side, because they were released thirteen hours apart and almost nobody does. American producer prices: +5.4%. Japanese producer prices, expected: +7.4%. The economy whose currency is being bought for its rate story has the cooler pipeline inflation. The economy whose currency is being sold has the hotter one, by roughly two hundred basis points.
The reason is not mysterious and we have published it before: on 8 September this desk cited Japan sourcing 95.9% of its crude from the Middle East in FY2024. An oil shock is mechanically worse for Japanese input costs than for American ones, and Japan cannot pass it into an export price without giving back the competitiveness the weak yen bought. So the same barrel that lifts US PPI by way of diesel lifts Japanese CGPI by more — and the currency effect runs the wrong way, because the market is trading the Fed’s reaction function and not Japan’s cost base. That asymmetry is real, it is quantifiable, and it is not a trade today. It is a reason to distrust any yen model that treats oil as a single global variable with one sign.
155 is a range boundary somebody published, not a target
Gaitame.com ran two notes this morning. The 07:55 JST piece by Tsutomu Nakamura of the Gaitame.com Research Institute gives a morning range of 153.600–155.300 and argues for a test of 155 on accelerating US inflation, high crude and rising Treasury yields, putting the market-implied Fed hike probability above 70%. The 08:00 JST market outlook, by Kawahata with DZH Financial Research data, frames the same setup as a recovery attempt toward the 9 September high and notes crude near $104.
Treat 155.300 as what it is: the upper bound of one house’s expected range, published at 07:55 JST, not a level where anything is known to sit. We have no option-expiry data for today’s New York cut — that page is a pre-US input on our source list, not an 08:00 JST one, and we are not going to invent strikes to fill the gap. What we can say is that our own published record has USD/JPY printing 152.89 on 8 September — the yen’s strongest against the dollar since February — so a test of 155 today would be roughly two hundred and ten pips of retracement in three sessions on a pair that had spent six sessions going the other way.
One number is worth carrying into the session. Our two sources disagree on where New York actually closed. Gaitame’s outlook puts the close around 154.50; Fisco’s New York summary, carried by Zaikei, has the pair reversing to 153.85. The high (154.67) and the low (153.85) are agreed by both. If your system needs a daily close to compute anything — a moving average, a range break, a gap — you are choosing between two published values sixty-five pips apart, and you should know which one your data vendor gave you.
A five-day trend that ends on day six is the exact shape that costs money
This is the sizing question, and it is uncomfortable. A trend that runs five sessions is long enough for almost every momentum filter to confirm it. Two-day breakout, three-day close filter, a 20-period channel, a slow-fast crossover — five consecutive lower closes clears all of them. Which means the systematic short-dollar position was at its largest on the evening of day five, immediately before the reversal, and the discretionary conviction was at its highest at the same moment for the same reason.
That is not bad luck. It is the mechanical consequence of confirmation lag, and it is priced into the expectancy of every trend system that has ever been backtested honestly. The response is not to add a filter that would have caught this one. The response is to know, in advance, what fraction of your annual drawdown budget a five-day trend reversing on day six is allowed to consume, and to have been sized so that the answer was survivable before you knew which day it was.
Concretely, for this session: the Japanese print at 23:50 UTC is a genuine event window in thin Tokyo liquidity, and it is a number nobody is positioned for, because a corporate goods index has never moved USD/JPY on its own. The dangerous version is not the print. It is the print landing on a book that was rebuilt at 154.50 twenty minutes earlier by a system that just re-entered long dollars on a one-day reversal — the mirror image of the day-five problem. If your re-entry rule is faster than your exit rule, you have built a machine that buys tops and sells bottoms and calls it responsiveness.
What this does not tell you
It does not tell you the yen rally is over. Six sessions of yen buying reversing in one session is consistent with a squeeze inside a continuing trend and with a genuine turn, and nothing published overnight distinguishes them. Friday’s CFTC Commitments of Traders release is the direct test — the yen net short was −92,227 contracts at 4 September, and we have been waiting on the update for a week.
It does not tell you where the New York close was. See above: our two sources give 154.50 and 153.85 and we are not adjudicating between them.
It does not give you today’s option expiries, because we do not have them.
It does not confirm one interesting claim we came across and are deliberately not treating as fact. Trading Economics attributes part of the upward pressure on US yields to “Japanese currency defense efforts prompting Treasury security sales by Tokyo” — the repatriation channel we have listed as an open question for four runs. That is a single vendor’s editorial framing on a quote page, not data, and we are recording it here rather than building on it.
And it does not tell you whether the Bank of Japan hikes on 18 September, only that FXStreet reported yesterday that markets fully price a 25bp move. That figure replaces the 97–98% we have been carrying since 8 September without a refresh, and it is the first time in four days we have re-read it.
Related
- FX events calendar — the Tokyo and London windows in UTC
- Signals
- EA presets
