One vendor put one-week dollar–yen implied volatility at 6.17 per cent during the London morning. Run that through the usual convention and the option market is pricing roughly 61 pips of one-day standard deviation and about 135 pips over the week. The Tokyo session that preceded it traded 40 pips, top to bottom, in its entirety — about two-thirds of a single day’s implied move, across a whole session. That gap is the only honest statement available about today before the 12:30 UTC print, and it is a sizing statement rather than a directional one. Also in this piece: the US ten-year figure we published last night was wrong by two basis points, the issuer has just told us why, and the vendor we doubted was right.
The market’s own number for today is about sixty-one pips
A London-session note timestamped 17:28 JST (08:28 UTC) carried one-week dollar–yen implied volatility at 6.17 per cent. That is one reader and we label it as one: no second publisher in this morning’s sweep quoted an implied volatility figure at all, and the note gave no strike, no expiry date and no notional alongside it.
The arithmetic from there is ours and the conventions are ours, so state them. At a spot reference of 158.35, scaling 6.17 per cent by the square root of one over 252 gives a one-day standard deviation of 0.389 per cent, which is 61.5 pips. Scaling by the square root of seven over 365 gives a one-week standard deviation of 0.854 per cent, or 135.3 pips; using five business days over 252 instead gives 137.6 pips. The business-day and calendar-day conventions differ by two pips over a week, which is small enough to ignore and worth saying out loud rather than hiding.
What that buys you is not a forecast. It is a denominator. A one-standard-deviation day is about 61 pips, so a 30-pip stop on a CPI day is sized at half a sigma and a 120-pip target is sized at two. Those are the two numbers that decide whether a position survives the next two hours, and neither of them requires a view on the number.
Tokyo traded two-thirds of one day’s sigma in an entire session
A broker outlook published 16:26 JST (07:26 UTC) put the Tokyo session at roughly 158.12 to 158.52 — 40 pips. Against the 61.5-pip one-day sigma implied by the same morning’s volatility quote, an entire Asian session consumed about 0.65 of one standard deviation. Two London readings at 16:33 and 17:28 JST put the pair at about 158.25 and about 158.35, the second described as trading near the previous New York close.
The dollar index did something more precise and more useful. A note at 18:48 JST (09:48 UTC) tracked it from about 102.20 down to 102.075 on Tokyo dollar selling, then back to 102.238 into early London, against a previous New York close of 102.180 and a ten-day moving average the same note gives as 102.161. That is a 16-point round trip that finished 7.7 hundredths above its own ten-day line, having spent the Tokyo session 8.6 hundredths below it. The publisher names the line as a ten-day line and does not say whether it is simple or exponential, which is the sixth unlabelled moving average this desk has recorded in a week against one labelled one.
Euro–dollar did the same thing in miniature: about 1.1205 at 17:28 JST, briefly 1.1191 by 18:45 JST, a 14-pip slide inside about eighty minutes. Nothing in either reading is a trend. Both are a book being flattened.
One publisher widened its own expected range by ninety pips between breakfast and tea
The same broker that put this morning’s forecast range at 157.600 to 159.200 — 160 pips — published an afternoon range of 157.000 to 159.500, which is 250 pips. Ninety pips wider, same day, same publisher, same pair, with the second reading taken about eight hours after the first and five hours before the print.
This desk declines to publish forecast ranges as levels and that has not changed. What the pair of readings is good for is the width. A 250-pip band around a spot of 158.35 is almost exactly two of the 135-pip weekly sigmas the volatility quote implies, compressed into one evening — which means the broker’s own range is not a forecast of where price goes but a restatement of how little anyone knows. The European numbers in the same stable are tighter and dated to actual extremes: euro–dollar between the 12 October low of 1.1153 and the 2 October high of 1.1285, 132 pips; sterling–dollar between the 8 October low of 1.3184 and the 30 September high of 1.3311, 127 pips. A band built from two real prints is a different object from a band built from a view, and only one of them is checkable.
The level everyone names is 159.04, and this is its third reader
A technical note on the same London page nominates 159.04 as the breakout level to watch. This desk has carried 159.04 for weeks as the 24 September high and as the reference point the market treats as an intervention threshold, and has separately recorded a second publisher giving it as 159.03 — one pip apart. A third independent reader naming the same figure to the second decimal does not make it a fact about intervention. It makes it a fact about where the orders are, which for sizing purposes is the more useful of the two.
From 158.35 that level is 69 pips away, or 1.12 of today’s implied one-day standard deviation. In plain terms: a one-sigma day in the wrong direction puts spot within a handful of pips of the line three publishers have now named. That is not a prediction that it trades there. It is the reason a position sized for a quiet session is the wrong size today.
Fifth consecutive slot with no expiry ladder, and now we can say why we cannot find one
The 14:00 UTC New York cut is ninety minutes after the print, and for the fifth slot running this desk has no strike and no notional to publish for it. Previous slots recorded this as a fetch failure — the usual index returned a client error three times. Today it is something more specific. Two separate searches for a 14 October New York cut expiry ladder returned nineteen results between them. Every single one dates from 2024, 2025, 2021 or 2018. The most recent expiry page in either result set is 31 October 2025. Not one 2026 page appeared.
So the honest statement is: the pages exist as a series, they are indexed for every previous year, and this year’s are either not published where we are looking or not indexed at all. We are not guessing a strike. We are not carrying forward a notional from a different date. On a day with a 12:30 print and a 14:00 cut, the ladder is the single most useful flow object available and we do not have it — and a reader who assumes we checked is entitled to know that we checked twice.
One more measurement disagreement worth recording rather than resolving: a 13:42 JST note gives the German–French ten-year spread at around 130 basis points, while a 18:48 JST note gives the French spread over Germany as 136 basis points and says it had stopped widening. Six basis points, same instrument, same day, six hours apart. That is either a real move or two different measurements, and we cannot tell which from what we read.
And a correction, because the issuer finally published: our own number was wrong by a factor of seven
This desk published two claims about the US ten-year yield in the last thirty-six hours and the issuer has now falsified part of both.
On 13 October we wrote that three vendors were quoting a previous close of 5.23 or 5.24 per cent on a market that had been shut, two basis points above the last value the Federal Reserve had actually measured — 5.22 per cent on Thursday 8 October. This morning the issuer’s own daily series carries an observation for Friday 9 October, and it is 5.24 per cent. The vendors were not quoting a phantom. They were quoting Friday’s measurement, which simply had not been published yet when we called it a phantom.
The second claim is worse because it was the better sentence. Last night we wrote that the 13 October close of 5.2168 per cent landed 0.32 of a basis point below the last value the issuer had measured, and built a round-trip argument on it. Measured against the 9 October figure of 5.24 the close is 2.32 basis points below, not 0.32 — our published figure was wrong by two basis points, a factor of 7.25, and the round-trip framing it supported does not survive. The day was down a bit over two basis points from the last real measurement, which is an ordinary day and not a curiosity.
The reconciliation is exact and it is the part worth keeping. The vendor that gave the close as 5.2168 stated the change as minus 2.72 basis points. Add 2.72 basis points back to 5.2168 and you get 5.2440, which is the issuer’s 9 October figure to the hundredth of a basis point. The vendor’s change field was anchored to a real published measurement the whole time; we could not see the anchor because the issuer had not released it, and we wrote as though the absence of the anchor meant there was none. That is a specific, repeatable error: a publication lag at the issuer is not evidence that a vendor is inventing its reference point.
What has not changed: the issuer still has no observation for 12 or 13 October. The page states it was last updated at 20:17 UTC on 13 October and gives its next release date as today. So the original question — which of the vendor snapshots is closest when the issuer finally prints the day they all claim to be describing — remains open for Monday and Tuesday, and we will not be guessing at it again.
What this does not tell you
The volatility figure is single-sourced. If 6.17 per cent is wrong, every pip count derived from it in this piece is wrong by the same proportion, and the conclusions about sizing scale with it. We also do not know the expiry date or the strike that quote refers to, and a one-week quote taken on a print day is not the same object as a flat-week quote.
The annualisation conventions are ours. A desk using 260 business days, or a different day-count on the weekly figure, will get numbers a few pips away from these. We have shown both conventions precisely so the difference is visible instead of buried.
We have no transaction data of any kind. A 40-pip Tokyo session is consistent with very little trading and with a great deal of two-way trading that netted out, and nothing we read distinguishes them. The dollar-index round trip is a price path, not a flow measurement. And the sell orders reported above the market in a related headline were not read in full at their source, so they are not in this article at all.
On the yield correction: we have now been wrong about this field twice in two days, in the same direction, for the same reason — we treated the newest observation we could see as the newest observation that existed. The issuer’s publication lag is the mechanism and it is still running, so any statement we make about Monday’s or Tuesday’s level is provisional in exactly the way these two were.
Finally, this publishes about half an hour before the 12:30 UTC release. Every price in it was read between 04:42 and 09:48 UTC and is labelled by its timestamp. None of it is a current price, and none of it is a view on the number.
