Four days ago this desk put a number in print that nobody had published. The Treasury was selling 119 billion dollars of three-, ten- and thirty-year debt across three days; the three-year came in at 58 billion and the ten-year at 39 billion; and we wrote that the thirty-year reopening at 17:00 today was therefore 22 billion dollars, flagged in our own run log as our arithmetic rather than a sourced figure, with a deadline of this slot to find a source or drop it. The source exists. It is 22 billion dollars. And in the hours while we were looking for it, a Federal Reserve governor stood up in Istanbul and published a December probability fifteen points above the one in our file.
The subtraction closed positive, and the issuer publishes the letter instead of the number
A preview filed at 11:22 on 5 October states the offering for today’s thirty-year as 22 billion dollars. A second reader, writing on 10 September about the previous sale, gives that one as 22 billion dollars too. Our subtraction was right, and it was right for a reason worth knowing: October is not a refunding month.
Go to the Treasury’s own tentative auction schedule and today’s entry reads “30-Year BOND R” — announced 1 October, auctioned 8 October, settling 15 October. The R is the reopening flag. Yesterday’s ten-year carried it as well; Tuesday’s three-year did not. That single letter is the field that determines the size, because reopenings in this cycle have been coming at 22 billion dollars while the new issues that cap a quarterly refunding have been larger. August’s thirty-year was a new issue of 25 billion dollars, the last leg of a 125-billion-dollar refunding announced on 5 August.
So the document that tells you which auctions are 22 billion and which are 25 billion is the one document that says, in its own text, that it carries no offering amounts and no times. A desk deriving the thirty-year size by subtracting the other two legs from a published three-day total would have been right in September, right today, and three billion dollars — 13.6 per cent — wrong in August. That is not a forecasting error. It is reading a calendar without reading its footnotes.
September’s reopening stopped through the when-issued, and the dealers barely showed up
Do the subtraction yourself before you call anything a tail. September’s thirty-year cleared at 5.308 per cent against a when-issued level of 5.335 per cent. The high yield came in below the pre-auction level, by 2.7 basis points. That is a stop-through, not a tail, and one of the two readers in our file labels it “tail: −2.7 basis points”, which is a phrase that means the opposite of the word it starts with. A second publisher, writing independently the same afternoon, puts the secondary thirty-year at 5.335 per cent ahead of the sale and the pre-auction range at 5.285 to 5.335 — the same number from a source that never mentions when-issued at all. The stop-through survives the cross-check.
The allocation is the part a system trader should care about. Bid-to-cover came in at 2.61 against a 2.38 average. Indirect bidders took 79.5 per cent against a 66.4 per cent average, 13.1 points above it. Primary dealers were left with 2.2 per cent of competitive awards against an 11.5 per cent average — under half a billion dollars of a 22-billion-dollar sale, a fifth of the normal share. For contrast, August’s new issue left dealers with 2.9 billion dollars, which is 11.5 per cent of 25 billion, identical to the stated average. One month the dealers took the average and the next month they took a fifth of it, at a clearing yield 9.2 basis points higher.
A governor put December at about 85 per cent this morning. Our file says 70.
At the Istanbul Economic Forum today, Governor Waller gave a speech published on the Federal Reserve’s own website in which he cited futures-market pricing of about an 85 per cent chance of at least one hike by the December meeting, and about 80 per cent odds of at least two hikes by March 2027. He added that sixteen of the eighteen participants who submitted September projections expected at least one more increase at the two remaining 2026 meetings, that core PCE rose 0.25 per cent month on month in August for a twelve-month rate of 3 per cent, and that twelve-month core inflation has sat between roughly 2.5 and 3.0 per cent since the spring of 2024.
Our own file has carried December at “about 70”. That is fifteen points below a sitting governor’s reading of the same futures market, delivered from a podium on the morning of a thirty-year auction. We are not going to resolve it by picking the higher number because it came from a committee member. Our 70 may well have been the probability of a hike at December rather than by December, in which case the two figures are not comparable and the fault is in our own labelling, which is a failure mode this desk has now documented in four other people’s tables and should have expected in its own. Either way, the operational instruction is the same one we have written before about central-bank probabilities: name the feed and the horizon in the same breath as the number, or do not carry the number.
The October reading is the cheaper check and it is also unresolved: the four feeds in our file put this month’s increase at 17, 20, 22 and 25 per cent, an eight-point spread, and a second reader this morning puts it at roughly 20 with next week’s CPI named as the catalyst. A pause on 27–28 October is the consensus position across every one of them.
Five points between one hike and two is a sizing statement, not a direction
Here is what we take from Waller’s two figures, and it is deliberately not a view on the dollar. At least one hike by December is priced at about 85 per cent. At least two hikes by March 2027 is priced at about 80 per cent. Those are different horizons and different event counts, so the five-point gap is not a conditional probability and we are not presenting it as one. But it bounds something useful: the market is not pricing much mass on a Federal Reserve that hikes once and stops. Whatever probability sits on “exactly one and done” across that overlap is small, and a second reader this morning has traders expecting at least three more increases by the end of 2027, which would put the range at 4.50 to 4.75 per cent.
If you are short yen on the policy gap, that is the asymmetry you are actually positioned in, and it is not the one the headlines describe. The headline risk is framed as whether the Fed hikes. The priced risk is almost entirely about the pace of a sequence the market already believes in. Those two risks want different exposures: the first wants event-window flattening around a single meeting, the second wants a carry position sized to survive a run of them. The ten-year is at 5.341 per cent this morning, a level one reader describes as unseen since 2002, and the thirty-year’s secondary yield was 5.61 per cent on 1 October against the 5.308 it last cleared at — 30.2 basis points of repricing since the September sale, on a single reader’s figures. That is the move you are financing.
The dollar–yen was 158.28 at 10:00, up about 0.1 per cent, having come 70.0 pips off a Tokyo low near 157.58 and still 23.0 pips short of yesterday’s Tokyo high at 158.51. The 17:00 auction is three hours after the New York cut and four and a half hours after the only scheduled American print of the day. Of those three windows, the auction is the one with 22 billion dollars of confirmed supply behind it and no consensus figure in front of it.
What this does not tell you
It does not tell you the thirty-year size for any month other than this one. Two readers give 22 billion dollars for today and one gives 25 billion for August, and we have not checked November, which is a refunding month and on this pattern should be larger. The reopening-versus-new-issue mapping to size is an inference from three data points, not a rule we have seen the Treasury state.
It does not give you a when-issued level for today. September’s stop-through is two sources agreeing on a number from a month ago. Nothing here tells you where today’s thirty-year is trading into 17:00, and the one secondary figure we have — 5.61 per cent on 1 October, with 5.49 on 25 September — is a week old and comes from a single reader.
It does not reconcile 85 and 70. We have stated the gap and named the most likely explanation, which is that our own figure was for the wrong horizon. We have not established that, and until we do, neither number should be load-bearing in a position.
And it does not tell you what the dealer share will be today. A 2.2 per cent takedown on a reopening and an 11.5 per cent takedown on a new issue, one month apart, is two observations. We have offered no mechanism linking the reopening flag to the dealer share, because we do not have one.
