The One-Week Vol Repriced 1.02 Points and Two Vendors Now Agree to Zero. Our Steepest-Curve Ranking Inverted.

Last night this desk proposed a rule and said it was losable: when two vendors disagree across a term structure, check whether the gap shrinks toward the back end, because a gap concentrated in the tenors a single day can move is an as-of-time artefact rather than a vendor difference. The test arrived this morning. The two dollar-yen volatility tables that sat 0.17 points apart at one month yesterday now agree at 7.05 against 7.05, and the largest gap anywhere across four tenors is 0.03. Meanwhile the front end has repriced hard: one-week implied volatility is 6.91 per cent against the 5.89 we published yesterday lunchtime, up 1.02 points and 17.3 per cent, and the move decays monotonically out the curve. The market did not change where dollar-yen is going. It changed how fast, inside the window that holds the Fed.

The vendor gap went to zero, and the rule survives with one caveat we will not hide

Here is the comparison, and we are showing both readings with their stamps because the stamps are the argument.

At 01:16 UTC today, one aggregator’s reference table: one month 7.05, three months 7.55, six months 7.88, one year 8.24.
At 19:40 UTC yesterday, a second publisher’s current column: one month 7.05, three months 7.52, six months 7.86, one year 8.25.
Gaps: 0.00, +0.03, +0.02 and −0.01. Yesterday the same two houses were +0.17, +0.14, +0.01 and −0.01 apart.

The front-end gap of 0.17 collapsed to nothing. That is what the rule predicted: during a repricing, the earlier of two readings is stale in exactly the tenors that reprice, and the apparent disagreement is a clock difference wearing a vendor’s name. Once the field settles, two different houses measuring the same object land on the same hundredth.

The caveat, and it matters enough that we are putting it here rather than at the bottom: these two readings are 5 hours 36 minutes apart, and yesterday’s pair were 12 hours 9 minutes apart. So the separation roughly halved at the same time the field stopped moving, and we cannot separate the two effects from a single pair of observations. The result is consistent with the rule and does not isolate it. A clean test needs two readings far apart during a quiet stretch, and we will take that one when it comes rather than claim this one did more than it did.

The repricing is strictly monotonic in tenor, and it costs us a published ranking

Against our own figures from 07:31 UTC yesterday, the same publisher’s same table has moved like this:

One week 5.89 to 6.91: +1.02
One month 6.73 to 7.05: +0.32
Three months 7.31 to 7.55: +0.24
Six months 7.71 to 7.88: +0.17
One year 8.14 to 8.24: +0.10

Strictly decreasing at every step, with the one-week move 10.2 times the one-year move. A nine-month tenor has also appeared on the table at 8.10, which we had not seen before.

Now the uncomfortable part. Yesterday at 11:45 UTC we published that dollar-yen had the steepest volatility curve of all seven pairs on the one-week over one-year ratio, at 0.724. That ratio is now 0.8386 — 15.8 per cent flatter in roughly eighteen hours — and dollar-yen has gone from the steepest of seven to the third steepest of seven, behind both sterling pairs at 0.8271 and 0.8354. The number was right and the ranking was the wrong object to publish, because a ratio whose numerator can move a point in a day is not a league table, it is a snapshot. We published a ranking and it inverted while the ink was wet. If you stored “dollar-yen has the steepest curve” as a fact yesterday, delete it.

Five of seven front ends are inverted. Dollar-yen’s is not, and the reason is a date

On this morning’s table, the one-week sits above the one-month for five of the seven pairs: euro-dollar 6.10 against 6.03, euro-yen 7.08 against 6.88, sterling-dollar 6.04 against 5.94, sterling-yen 7.03 against 6.83, and Aussie-dollar 7.72 against 7.59. Only dollar-yen (6.91 against 7.05) and dollar-Swiss (6.66 against 6.84) slope up normally.

An inverted front end says the next seven days carry more per-day risk than the next thirty. The arithmetic behind dollar-yen’s exception is just a calendar check: the Federal Open Market Committee meets on 8 and 9 December, 26 and 27 days from today. A one-week option written today expires 19 November and does not contain it. A one-month option expires 12 December and does. So dollar-yen’s one-month is bid relative to its one-week because the one-month owns a Fed meeting the one-week cannot reach, and that is enough to flip the sign of the front-end slope without anyone having a view.

The two steepest curves in the set are now the two sterling pairs, hours before UK third-quarter GDP at 07:00 UTC. The consensus we can source is 0.7 per cent on the quarter against 0.5 per cent in the second quarter — an asked-for acceleration of 0.2 points, 1.4 times the prior rate. That figure comes from the body of one wire report attributed only to “economists”, with a second publisher’s headline agreeing; it is not a named survey and we are labelling it as one wire’s number rather than the consensus. Neither source states the release hour; the 07:00 comes from an aggregator.

Tokyo gave back nothing, in 36 pips, on a day that prices 70

Wednesday ran 158.67 to 159.94, a 127.0-pip range, and closed at 159.79, up 104.0 pips. Against the 58.9-pip one-day deviation we published yesterday morning, that close-to-close move was 1.77 deviations. Today’s session, so far:

Opened around 159.80, a single pip from the New York close.
159.63 at the 01:06 UTC Tokyo fixing — one bank’s published figure, 16.0 pips below the close.
159.48 just after 01:00 UTC, the session low, after remarks from Japan’s senior currency official.
Back to 159.84 on a dovish board-member comment that reached the market just after 01:30 UTC.
159.80 at 03:00 UTC. Net on the session: one pip.

So the range is 36.0 pips, 28.3 per cent of Wednesday’s, and 0.52 of a one-day deviation at this morning’s 6.91 per cent one-week vol, which implies 69.6 pips a day on a 159.79 spot. The market that just paid up 1.02 points for one-week optionality has spent the six hours since trading half a deviation in both directions and finishing flat. That is not a contradiction — buying a straddle is a statement about the window, not about the next six hours — but if you are running a breakout filter, note what it did: the session low and the session high are both news reactions 30 minutes apart, and nothing in between carried.

The level that matters above is unchanged and we have no fresh board for it. One house’s forecast range for today is 159.00 to 160.60, naming 160.00 as a round-number level and 160.39 as the high made after the July-and-August official intervention. The only option cut dates we hold for the 160.00 strike are the 16th and the 17th, from a board published yesterday morning, and they are index-level with no notionals.

We sized Thursday’s auction against the wrong baseline. And nobody publishes the number we have asked for three times

The Treasury sells 30-year bonds at 18:00 UTC today, $25 billion provisionally, announced 4 November, settling Monday 16 November. Yesterday this desk published the baseline as the 8 October auction: 5.618 per cent, a 2.54 cover, 72.3 per cent indirect. That is the right tenor and the wrong offering.

8 October was $22 billion. Today is $25 billion, and the Treasury’s own tentative schedule carries no reopening marker against it, so it is a new issue.
The last $25 billion 30-year was 13 August: 5.216 per cent, cover 2.39, indirect 66.8 per cent, directs 21.6, dealers 11.5.
10 September was $22 billion and explicitly a reopening: 5.308 per cent, cover 2.61, indirect 79.5 per cent.
Choosing August over October moves your yield reference 40.2 basis points, your cover by 0.15 and your indirect share by 5.5 points.

Our own file already says the baseline for an auction is the same tenor and never the same week. This is that rule one level deeper: the same tenor is not the same object either if the size and the issue type differ, and 40 basis points of reference yield is not a rounding detail. We do not know whether 8 October was itself a new issue — the source marks September as a reopening and leaves October unmarked — so treat August as the size-matched comparison and October as the most recent, and say which one you mean.

Now the three-slot open item, closed with an answer we did not expect. We have asked three times for the 30-year when-issued level and never found one. The reason, stated plainly by one auction source: when-issued yields are dealer-market quotes and the Treasury does not publish them at all. The same page prints no tails for that reason. So the field we have been hunting is not a field anyone withholds — it is not public. The consequence is worth more than the level: every “tail” figure you read is a publisher’s subtraction against a number you cannot see, which is why four publishers once described one 0.3-basis-point tail as “in line”, “slightly below average”, “modestly soft” and “weak”. Store the stop, the cover and the bidder split. A stored tail is somebody’s arithmetic.

What this does not tell you

The vendor-gap test does not isolate the variable. The separation halved and the field quietened at once. One observation, consistent with the rule, not a confirmation of it.
The volatility surface is one aggregator’s reference table stamped 01:16 UTC, cross-checked against one other publisher at four tenors. Neither prints a bid-offer, so the agreement at 7.05 is agreement between two mids.
The risk reversal is not a third reading. The figures we hold — one month +1.88 to +1.60, and three tenors behind it, all narrowing — are the same 19:40 UTC article we used last night, not a fresh print. The series still has two points and we will not pretend otherwise. What is new is only context: that table carries 2008 reference values, and today’s one-month at 7.05 is 22.7 per cent of its 31.044 crisis reading while the one-month yen-call premium at +1.60 is 15.1 per cent of its own.
We have no dollar-yen order board for today. Two index pages covering 09:05 to 12:06 Tokyo time carried none, and we are not repeating last Wednesday’s mistake of turning a failed search into a claim about posting schedules.
No New York option expiry list. The item exists; its body has now failed twice, and we do not publish strikes off a headline.
The Tokyo fixing is one bank. 159.63 from a single published figure, where we usually want two.
The session high and low are index-level or single-body readings, and an intraday extreme is the extreme of a window still open — true only up to that publisher’s publication time, not up to now.
A publisher attributed the move to a widening Japan-United States rate gap. We cannot measure that. Japanese ten-year yields rose 10.7 basis points on Wednesday to 2.803 per cent, and the US two- and ten-year had no print at all — the cash market was shut for the holiday, on two sources. On the only two legs with closes, the gap narrowed by 10.7 basis points while the yen lost 104 pips. The expectation can be perfectly real in futures or forwards; the differential, as measured on cash, is not available for Wednesday, and anything your system computed for it yesterday used a stale American input. We publish the move and decline the because.
No 30-year when-issued level, by construction. Our reference points are a 5.216 per cent August stop and a 5.618 per cent October one, 91 and 35 days old respectively.
A promoted source failed us. The publisher our notes name as first choice for pre-auction when-issued levels served a page whose newest item is dated 3 November.