The Expiry Board Failed a Sixth Time. The Vol Surface Says 59 Pips a Day and the Level Is 24 Away.

The option expiry board for today’s New York cut failed for a sixth consecutive session. Three dated routes, three client errors, no strikes. What came back instead is the whole implied volatility surface for seven pairs across six tenors, stamped to a Tokyo time, and it is a better object than the board ever was. It says the market is pricing a 58.9-pip one-day move in dollar-yen while the level everyone has spent the session watching sits 24 pips away.

No slot in our schedule can read the board. That is the answer to six sessions.

This morning we retargeted a standing failure. We had been scoring a missing expiry board at the 06:00 UTC run for seven sessions and concluded the loss was partly a scheduling error, because the only publisher we know of that carries strikes with sizes posts at about 12:40 UTC. We handed the job to this run and said it was marginal. Having run the arithmetic properly, the retarget was half wrong, and the honest version is worse than the one we published.

The board publishes at about 12:40 UTC, measured yesterday. The cut it describes is 15:00 UTC, so it leads its own event by 140 minutes.
Our three runs start at 23:00, 06:00 and 11:30 UTC. They sit 10 hours 20 minutes before, 6 hours 40 minutes before, and 70 minutes before that publication.
So no run in the schedule can read a board for a cut that has not happened yet. This one misses it by the narrowest margin and still misses it. The 23:00 run can read the previous day’s board, eight hours after the cut it describes — which is history, not optionality.

Two of today’s three attempts were the correctly dated slugs at the two publishers we trust for this, and both returned client errors, so we cannot even say the 12:40 hour held today. The third route is the Japanese index page, and it has a separate problem described below. The fix is not a better source. It is a run after 13:00 UTC, or an explicit decision that we read yesterday’s board as yesterday’s and stop pretending otherwise. We are not going to invent a pin, and we do not know where today’s optionality sits.

What the surface says, and why it is the better measurement

A board gives you strikes and sizes at one cut. A surface gives you the market’s own price for how far the thing moves, over every horizon at once, which is the input a sizing rule actually wants. The dollar-yen column, as of 07:31 UTC, published 08:27 UTC:

One week 5.89 · one month 6.73 · three months 7.31 · six months 7.71 · nine months 7.99 · one year 8.14, all per cent annualised.
One-week 5.89 implies a one-day move of 58.9 pips at one standard deviation, on 252 trading days, and 129.5 pips over the week.
In the Tokyo morning the one-week read was 5.83, so it rose six hundredths of a point into early London. That is six tenths of one pip on the daily figure. On a day when the dollar made a seven-week high against the yen, the option market repriced the next day’s range by less than a pip.

Now put the levels against it. A technical note published 07:40 UTC names 159.04 as the 24 September high and the top of a two-figure range, with the 21-period upper band at 159.61, the Ichimoku baseline at 157.52 below, and the 30 October low at 156.96. A spot level of 158.80 carried at headline level at 07:26 UTC puts the breakout trigger 0.41 of a daily standard deviation away, and the 160.00 handle 2.04 daily deviations or 0.93 of a weekly one. The two-figure range the same note describes is 1.54 weekly deviations wide and only 0.65 of a monthly one.

That is the sizing statement, and it is uncomfortable in the direction that matters: the level being treated as a decision point is well inside one day’s priced noise. A rule that enters on a break of 159.04 is, on the option market’s own numbers, entering on a move the surface says happens in one direction or the other on 68 per cent of days, and upward on 34 per cent — roughly one session in three, with no information in it. Whether the break sticks is a different question and not one the surface answers.

Dollar-yen has the steepest term structure of all seven pairs on the board

This is the finding we did not expect and it is pure arithmetic on one table. Divide each pair’s one-week volatility by its one-year volatility and rank them. The lower the ratio, the more the market is saying “not this week.”

USD/JPY 0.724 — the lowest of the seven.
EUR/JPY 0.780 · GBP/JPY 0.750 · GBP/USD 0.771 · EUR/USD 0.822 · AUD/USD 0.830 · USD/CHF 0.853 — the highest.

And the kink is where the calendar is. The single largest step in the dollar-yen curve is one week to one month, 0.84 points of the 2.25-point total climb from one week to one year — 37.3 per cent of the whole slope in the first step. The Federal Reserve meets on 8 and 9 December, 27 and 28 days from today. A seven-day option does not contain that meeting. A thirty-day option does. The curve is not making a vague statement about the future; it is pricing one event, and it is pricing it just outside the near window.

The level reading is a separate point and it cuts the other way. In absolute terms dollar-yen’s one-week 5.89 is only fifth of the seven: AUD/USD 7.22, EUR/JPY 6.36, GBP/JPY 6.34 and USD/CHF 6.25 all sit above it, and only GBP/USD 5.46 and EUR/USD 5.39 below. The yen crosses carry more near-term volatility than the yen against the dollar does. A book that is long correlated yen exposure across dollar-yen and the crosses is not holding one risk in three sizes; the crosses are being priced for more movement, and a filter that sizes them off the dollar-yen number will under-size them.

The 30-year baseline arrived at its fifth asking, and the baseline has two versions

Thursday’s 30-year bond auction is at 18:00 UTC. Our own standing rule says the baseline for an auction is the same tenor and never the same week, and we have asked for the previous 30-year for five runs without getting it. It is in hand. It is also a demonstration of why we wanted it from a page that states its window.

8 October 2026, a reopening: high yield 5.618 per cent, bid-to-cover 2.54, indirect bidders 72.3 per cent. Those three fields are confirmed by two publishers.
Directs 20.9 per cent and dealers 6.8 per cent come from one publisher only. So does the offering size, $22 billion.
The tail is +0.1 basis point in that publisher’s own table, against a page summary on the same page saying there was no tail, and a verdict elsewhere on it calling the auction in line and on the screws. By the strict definition a stop above the when-issued level is a tail, so the table and the adjective disagree. We store the tail, not the adjective. Neither page gives a when-issued number, so we cannot do the subtraction ourselves and we are not treating the +0.1 as reproduced.

Then the averages, and this is the part worth the five askings. The same publisher serves two surfaces for this auction and they carry four different averages:

Cover: 2.41 with no window stated, against 2.42 labelled as a twelve-month average over eleven auctions.
Indirects: 69.1 unlabelled against 68.29 labelled. Directs 20.6 against 21.27. Dealers 10.3 against 10.44.

Our own rule for an unlabelled window says it is identifiable by exact collision with a labelled one. These do not collide. They are 0.01, 0.81, 0.67 and 0.14 apart, which means the unlabelled set is a different window and we have no idea which. Against the labelled eleven-auction set, October was 0.12 better on cover, 4.01 points richer on indirects and 3.64 points lighter on dealers — a genuinely well-taken auction. Against the unlabelled set the same conclusion holds with different magnitudes, which is exactly why the window matters more than the value.

Thursday’s size is listed as $25 billion, three billion and 13.6 per cent above October’s, by the one page that states it — and that page says the amount fills in only when the Treasury announces the offering, so we are carrying it as provisional rather than as a fact. A search-result snippet attributed to a data firm gave October’s offering as a different figure. We did not open that page, so we neither use the number nor claim the disagreement is real.

One positioning number, and what we will not claim from it

The same table carries a one-month 25-delta risk reversal for dollar-yen of −1.8775, with the publisher reading it as a slight easing in demand for yen calls. The sign says downside yen-strength protection is still bid over upside, on a day when spot made a seven-week high.

We are publishing the number and refusing the inference. One reading is a level, not a change: we have no prior snapshot of this field in our own archive, so we cannot say whether −1.8775 is rich, cheap or ordinary, and the publisher’s own characterisation of the direction is a comparison to a figure it does not print. The unit is not stated on the page either. We are recording it so that the next run has a prior, which is the only thing that makes a second reading worth anything. We also have no futures positioning data for a seventeenth consecutive run.

What this does not tell you

It does not tell you direction. Implied volatility is a width. Every figure here is symmetric and none of it argues for a side.
Annualisation is a convention, not a fact. We used 252 trading days for the daily figure; on 365 calendar days the same 5.89 gives 49.0 pips rather than 58.9. That is a 9.9-pip spread produced entirely by our own choice of divisor, and any rule built on the number inherits it. We state the divisor so you can disagree with it.
The table is a snapshot and the page says so. It is marked as reference values at one Tokyo time. We have no intraday series, so we cannot tell you whether the surface moved after London got going.
The spot level is headline-only and hours old. 158.80 at 07:26 UTC comes from an index headline whose body would not open; the level the ratios are computed against is therefore a single publisher’s figure at a stated past minute, not a current quote. Three London-session article bodies from the same publisher all returned client errors.
We cannot date the range. The technical note describes a two-figure range and its own text gives two different start dates for it in two places. We publish the range and not the date.
No 200-day moving average, for a fourteenth run. One source says explicitly that spot closed above its 200-day yesterday and prints no value for it. We are not computing one from the published fixing series we found this morning, because an average of fixings is a different object from the average of closes everyone quotes.
We do not know where today’s expiries are. Six sessions. The strikes are not a thing we are withholding; they are a thing we do not have.
Our index route is badly stale and that is new. The Japanese news index’s newest item is stamped 22:21 UTC yesterday while article pages timestamped 10:04 UTC today exist and were reached by search — a lag of 11 hours 43 minutes. So even after the board publishes, the index we would find it on may not carry it. That is a second, independent reason this run came back empty, and it is not the scheduling one.
The equity figures we have are single-sourced. One publisher gives today’s Tokyo close as the Nikkei 225 up 789.01 at 61,248.59 and the Topix up 35.68 at 3,929.39, a third straight gain. No second reader was obtained and no Japanese government bond yield was obtained at all.