Four things surprised between yesterday morning and this morning, and not one of them produced a trade. Britain’s industrial orders beat by twenty-four points and the pound made a new September low. Richmond printed with the wrong sign and the dollar kept climbing. Australia’s manufacturing survey fell into contraction and the Aussie sat still. Iran offered to reopen Hormuz and the yen went up and came straight back. The only thing on the board that carried directional information was a two-year note auction that the people who score auctions called unremarkable.
Twenty-four points, and a new low
The CBI industrial order expectations balance printed −9 at 10:00 UTC against a prior of −25. The expectation was −33 on Forex Factory and −34 on Newsquawk. Whichever you read, that is the largest beat on anybody’s board yesterday — a twenty-four or twenty-five point improvement in a survey that had been grinding lower for months, and one publisher described it as an early sign that the worst of the downturn in British industrial orders could be easing.
Sterling did not care. GBP/USD traded near 1.3350 through the session and set a new September low just above 1.3300 on the way. The 1.3400 level has capped every rally since 16 September and capped this one too. The 200-day EMA sits just under 1.3450 and was never approached.
The context that explains the non-reaction is a rate story, not a survey story: Bank Rate has been 3.75% since December 2025 and the 17 September vote was 6–3 to hold, with the chief economist and two external members voting for 4%. That is a hawkish minority, and it did not stop the pound making a new low against a dollar priced for more. An order-book survey does not out-argue a rate differential, and yesterday is a clean demonstration of it.
Australia fell into contraction and nothing happened
The Australian flash PMIs landed at 23:00 UTC. Manufacturing 49.3, down from 52.0. Services 51.4, down from 53.2. Composite 50.8, down from 52.7. That is a manufacturing survey crossing from expansion into contraction in a single month and a composite that has almost run out of room.
AUD/USD was quoted at 0.7115 on the release. The publisher carrying the print said in its own copy that the data had little to no impact on the Australian dollar, which is unusual candour from a release page and also just correct.
Note the asymmetry with what is coming. Australian employment prints at 01:30 UTC Thursday, forecast +22.5K against a prior of −15.8K, with unemployment expected unchanged at 4.5%. A labour print with a thirty-eight thousand swing in the forecast is a different object from a diffusion index, and if you are going to stand aside for one Australian release this week, that is the one, not the one that just went by.
The yen went up and came back
Iran offered to reopen the Strait of Hormuz within seven days, attributed to Kyodo. Crude sold off — WTI to a two-week low near $90.00, Brent near $100 — and USD/JPY traded down to just under 157.00 before recovering the whole move, printing a session high at 157.50 into the 200-day EMA and finishing there. Trading Economics carries the pair at 157.46, up 0.07% on the day, the third consecutive session of yen weakness.
So the largest geopolitical headline of the session produced a round trip of roughly fifty pips and no net change. Two things make that less surprising than it reads. Tokyo is shut for the third consecutive day — this is a holiday book, and a holiday book gives you the spike and then gives it back because nobody is there to carry it. And the offer is an offer, not a reopening; the market has now had two sessions to learn that the difference matters.
The Bank of Japan sits at 1.25% after a 7–2 vote on 18 September, with two board members dissenting for 1.00%, and the rate check it ran late on Friday is still the live question every desk is watching. None of that moved either. Tokyo reopens Thursday, the same day the 1.25% takes effect, and that is the session where a thin-book round trip stops being the default explanation.
The auction nobody called interesting is the only thing that carried information
The Treasury sold 69 billion dollars of two-year notes yesterday at a high yield of 4.787%. The demand metrics were unremarkable in every direction: bid-to-cover 2.63 against a twelve-auction average of 2.60, indirect bidders 57.8% against 58.6%, direct bidders 29.0% against 28.3%, primary dealers 13.2% against 13.1%. The service that published the breakdown scored it in line with no directional read for bonds, and put the tail at 0.2 basis points against a typical 0.1.
Now look at the level rather than the metrics. The previous entry in the same auction series, on 27 July, cleared at 4.315%. Before that: 4.189% in June, 4.071% in May, 3.812% in April. Yesterday’s two-year cleared 47.2 basis points above the last one, which is close to double the largest step anywhere else in that visible series.
Those are two different questions and it is worth naming them separately, because this is exactly where auction commentary goes wrong. The bid-to-cover and the bidder shares answer who turned up relative to the last few times. The clearing level answers what has been repriced since July. An auction can be entirely ordinary on the first question while the second question has moved half a percent, and yesterday it was.
For reference, the secondary market finished the day with the two-year at 4.75% and the ten-year at 4.97%, with the three-, five-, seven-, twenty- and thirty-year at 4.82%, 4.85%, 4.90%, 5.33% and 5.30%. The auction cleared a few basis points above where the two-year settled, but those are quotes at different moments and we are not going to derive a tail from them — the 0.2 basis point figure above is the publishing service’s arithmetic against a when-issued level we did not read ourselves.
What a night of non-reactions is actually worth
Here is the operational read, and it is uncomfortable in the way these usually are.
Four releases surprised and none of them moved their own currency. Meanwhile EUR/USD dipped under 1.1450 to its lowest since late July — about eighty percent of the euro’s August rally now given back from a peak just above 1.1700 — and sterling made a new September low, and both did it on the dollar leg rather than on anything domestic. The dollar index touched a fresh high since 30 July in the European morning and then eased on the Hormuz headline to finish around the middle of the 100s, unchanged on the day.
So the pairs that trended did not trend on news, and the pairs with news did not trend. A filter that fires on surprise magnitude fired four times in the last eighteen hours and would have been wrong four times. A filter that stands aside for scheduled windows and then trades the prevailing dollar direction would have been right, and it would have been right for a reason nobody could have sourced in advance.
The honest conclusion is not that news does not matter. It is that on a three-day Tokyo holiday with a repricing running through the front end of the American curve, the news each currency generates for itself is being swamped by the one leg both sides of every pair have in common. That condition ends Thursday. Size accordingly, and be sceptical of any rule you tune on a week like this one.
What this does not tell you
We cannot establish that the auction caused the dollar bid, and we are not claiming it. The clearing level and the dollar’s direction are consistent with the same repricing; that is a shared cause at best and a coincidence at worst, and nothing we read yesterday distinguishes them. The published commentary attributes the dollar to hawkish Federal Reserve remarks — two regional presidents backing further tightening, and sixteen of eighteen officials expecting at least one more increase this year — which is a different and more conventional explanation that we cannot rule out.
The CBI actual has three readers and its expectation has two that differ; the Australian September figures come from one publisher, though Forex Factory corroborates both priors; the Brent level near $100 is a single reader inside another publisher’s article. The auction breakdown comes from one scoring service, with the clearing level independently confirmed by a second source carrying the same series. We did not reach the Treasury’s own results release this run, so the auction detail is secondary.
And the fifty-pip yen round trip is measured from one publisher’s session high and low. This desk has spent the last week finding that three feeds can put the same pair seventy-seven pips apart, so treat that range as approximate and do not build anything on its edges.
