Yesterday at 11:51 UTC this desk published that the issuer had no ten-year value for 14 or 15 October, and that the only route to a number was an area figure of 5.30 less a move of twenty-one basis points, which landed on 5.09. The issuer has now published the fourteenth. It is 5.19. We were ten basis points low, and the direction of our error is the interesting part: we reconstructed a fall of twelve basis points from the last published value, and the actual fall was two. Four hours before London opens, the fifteenth is still missing, today is the issuer’s stated release date for it, and the man who sets the price of money spent this morning on a stage in Bangkok declining, in terms, to supply a number of any kind.
We derived 5.09 for a day the issuer now prints at 5.19
The series read at 06:14 UTC this morning gives 14 October at 5.19, 13 October at 5.21, 12 October as a dated row with no value in it, 9 October at 5.24 and 8 October at 5.22. The last-updated stamp is 15 October at 3:16 p.m. CDT, which is 20:16 UTC. The stated next release date is 16 October — today.
So the fourteenth has arrived and it is 5.19. Our published reconstruction was 5.09. The error is ten basis points, and we would rather state the mechanism of it than the size. We built the figure by taking a vendor’s description of the level as “the 5.30 area” and subtracting a move that the same page described as twenty-one basis points in one place and twenty in another. Both of those inputs are prose. Neither was a measurement. The issuer’s own number says the fall from 13 October was two basis points, not twelve, which means the bulk of our error was in the starting point rather than the subtraction: “the 5.30 area” was describing something our arithmetic then treated as a close.
There is a further ambiguity we should own rather than paper over. Our sentence did not say unambiguously which day 5.09 referred to. Written on the fifteenth about a page carrying neither the fourteenth nor the fifteenth, it reads most naturally as the fourteenth, and we are scoring it as the fourteenth. If a reader took it as the fifteenth, the claim is still unscored, because the fifteenth has not been published. A figure whose referent date is a matter of interpretation is not a figure.
The two vendor levels for the fifteenth sit thirteen to sixteen basis points below it
Against the issuer’s 5.19 for the fourteenth, we hold two vendor quotes for the session that followed: 5.035 from a Tokyo desk note stamped 03:42 JST on the sixteenth, and 5.06 from an American wrap stamped 13:38 UTC on the fifteenth. Those are fifteen and a half and thirteen basis points below the issuer’s fourteenth, and two and a half basis points apart from each other.
Two readers that close together are a reasonable second source for each other. What they are not is a value for the fifteenth, because neither is the issuer and because one of the two carries a timestamp whose relationship to the New York close we cannot establish. The honest statement of where the American ten-year is this morning is therefore: the last published official value is 5.19, dated 14 October, and two secondary readers put the following session somewhere between 5.03 and 5.06. We are not splitting the difference and we are not calling 5.05 a level.
If the 15 October observation appears today as the issuer says it will, that becomes scoreable within hours. We are publishing the vendor pair now, before the fact, precisely so it can be scored rather than quietly forgotten.
The chair had the last word before the blackout and used it to give no number
Kevin Warsh appeared at 03:30 UTC in a fireside conversation with the managing director of the International Monetary Fund. By at least one reading it was his final public appearance before the committee’s pre-meeting blackout. He said he does not believe in prejudging meetings, that he would not say what he might do for the rest of 2026, that the world has broadly conformed to the committee’s expectations since September, and that judgements would be made at the October and December meetings. He called a structural investment surge the most important thing happening on the upside of the global economy. He noted that three-quarters of the institution’s staff do not work on monetary policy, and announced a field hearing on artificial intelligence in the San Francisco Bay Area on 16 November.
Not one of those is a rate, a probability, a path or a threshold. This matters to us because of a count we have been keeping: across nine officials checked over the preceding week, exactly one had cited a probability figure, and that citation was seven days old and superseded by the time we logged it. The chair, with the last word available to him before a blackout and an interviewer who could have asked anything, makes ten, and he is the strongest possible negative instance — not an official who happened not to be asked, but the one person whose number would have settled the question, declining it on principle in front of the fund.
The operational reading is unglamorous and we will state it plainly. If your process has a rule that upgrades or downgrades conviction on central-bank speech content, that rule has had ten consecutive opportunities this fortnight to find a numeric input in official remarks and has found one, stale. Speech-driven repricing in this cycle is being produced by the market’s interpretation of qualitative language, not by officials publishing figures. Those are different signals and they justify different position sizes.
Bangkok, not Singapore — and the reader that said Singapore is the one that corrected it
Yesterday evening we held two accounts of where this appearance would happen. One placed it at the fund’s annual meetings in Singapore; the other described a fund dialogue and named no city. We cut the venue from both of yesterday’s articles because two readers gave two answers and the location was not load-bearing for either piece.
This morning three independent readers place it in Bangkok, and one of the three is the publisher that said Singapore sixteen hours earlier. We record that as a self-correction rather than a conflict, and we note without sarcasm that the corrected version is the one that now carries a precise venue description. A publisher that revises a factual field and publishes the revised version is more usable than one that never commits to the field at all.
The reason to spend a paragraph on a city is that the venue determines the local hour, the local hour determines who is awake, and this appearance landed inside Asian liquidity rather than in front of a New York desk. The market reaction we can see is thin: one page puts the dollar index down 0.22 per cent at 101.82 with no timestamp on the quote, and dollar–yen is described as giving up 158.00 during the Asian session. We are not going to build a causal claim out of an untimed index quote.
What this hands London
Three things, none of them directional. First, there is no current official price for the American ten-year, and the two unofficial ones sit thirteen to sixteen basis points below the last official one. If a yield differential is an input to how you size yen risk this morning, it is an input you cannot presently source to an issuer.
Second, the policy question that moved this market all week has just been formally refused an answer by the person best placed to give one, with the next scheduled opportunity being the meeting itself on 27 and 28 October. Between now and then the only numeric inputs will be data and pricing, not speech.
Third, the European morning does carry a release block — the euro-area September final inflation set, measured against a flash of 3.8 per cent annual, up from 3.2 in August, with energy at 18.8 against 14.3 — and a final is a revision rather than a surprise. Treat a revision window as a liquidity event rather than an information event, size it accordingly, and keep the risk budget for the American industrial production print at 13:15 UTC and the positioning report at 19:30. The Bank of Japan governor’s speech, read on his behalf by the deputy governor, falls at 06:30 UTC, half an hour before the London open and after this piece was written.
What this does not tell you
It does not tell you where yields or the dollar are going, and we hold no directional view in this article. The 5.19 is one reading of one issuer page taken at one moment; we have not cross-checked it against a second official source, because for a constant-maturity series there is not usefully a second official source, and that is a limitation of the claim rather than a strength.
Our scoring of our own 5.09 against 5.19 rests on an interpretation of our own sentence, which we have set out above so that a reader can disagree with it. If you read that sentence as referring to the fifteenth, then we have not been proved wrong today — we have been shown to have written a claim that cannot be scored, which is a different fault and not obviously a smaller one.
We did not watch the Bangkok appearance; everything in section three is drawn from two written accounts published within five minutes of each other, and a speaker can say something numeric that neither write-up carries. The count of ten officials is our own tally from our own archive and is only as good as our sweep — a probability cited by an official we did not read would not appear in it. The dollar index quote has no time attached and we have used it only as colour. We have read no futures tool directly at any point this week, and every probability figure in our recent archive is a vendor’s reading of one.
Related
- FX events calendar — the day’s release windows in UTC.
- Signals — how we treat a speech window with no numeric content.
- EA presets — news-window and volatility filter settings.
- We published 43.7 per cent; seven minutes later the same feed read 31.6 — where the 5.09 reconstruction was published.
- We said the issuer was silent; it had published three hours earlier — the first instalment of this same correction.
