At 11:42 UTC yesterday this desk published that 78.4 per cent of a 6.979 billion dollar expiry board sat in the fifty pips between 157.50 and 158.00, that spot was inside that band at 157.76, that the largest single strike was 2.1 billion dollars at 157.85 nine pips above it, and that a pin looked likelier than a break. Two hours and eighteen minutes later the 14:00 UTC cut arrived. One reader puts dollar–yen at 157.86 at 20:07 UTC — one pip above that strike. Three readers put the day’s high between 157.86 and 157.89; not one of them has it at 158.00. The pin scored. What it does not do is tell you why, and that gap is the useful part of this note.
The marking, pip by pip
The claim was a band, a spot inside it, and a direction of error. All three hold. The pair spent the session inside the fifty pips carrying 5.469 billion of the 6.979 billion board, the high stopped eleven to twelve pips short of the 158.00 strike where 1.5 billion sat, and the close printed within a pip of the 2.1 billion at 157.85. The session range on the readings we hold runs from 157.42, taken at 10:20 UTC before the cut, to 157.89 — forty-seven pips, against a dense zone fifty pips wide and shifted eight pips above the low.
Now the discipline. A one-pip close against a two-billion-dollar strike is the kind of coincidence that makes a mechanism feel proved, and this desk published a near-identical coincidence yesterday — a Tokyo fixing landing exactly on the previous day’s high — and refused to build on it. The same refusal applies here, harder, because the stakes are higher. One observation is a data point. The honest version of what we got right is narrower than the headline: we said the notional was concentrated, that spot sat inside the concentration, and that the asymmetry favoured containment over a break. The asymmetry was the call and the asymmetry paid. The pip is decoration.
There is also a thing we cannot score, and we flagged it in advance: we have no reading of where spot actually printed at 14:00 UTC. The 157.86 is six hours after the cut. The close being inside the band is consistent with a pin and does not establish one, and a pair that wandered out and back between 14:00 and 20:07 would look identical in our file. Anyone who wants this scored properly needs a tick at the cut, which nothing we read publishes.
A moving average at 157.89 and a strike at 158.00, eleven pips apart
Here is why the pin is not a finding about options. One reader’s technical ladder, published 11:05 UTC, puts the 21-day moving average at 157.89 and lists 158.00 immediately above it, then the upper boundary of the Ichimoku cloud at 158.44 and the 200-day at 158.54. The day’s high was 157.88 to 157.89.
So the high stopped on a moving average, to the pip, and eleven pips under the option strike. Two explanations, one tape. If you went into yesterday believing option gamma pins price, the close one pip from the biggest strike confirms you. If you went in believing a 21-day average caps a grinding uptrend, the high landing on it to the pip confirms you instead. Both readings are available from the same forty-seven-pip session and nothing in it separates them — which is the honest reason we are not upgrading an expiry-concentration argument into a rule. When two objects sit eleven pips apart and price stops between them, the session has told you about the cluster and nothing about the cause.
A smaller corroboration, since it costs a line: our own note had the 200-day at 158.52 and this reader has 158.54. Two pips apart, two publishers, which is about as good as this file ever gets on a long moving average. Supports below: a 10-day at 157.79, an Ichimoku baseline at 157.56, the conversion line at 157.34, and 157.00 — where a large 29 October cut still sits, so the yen board’s weight is parked on European Central Bank day and then on Bank of Japan day, not on today.
Five billion Australian dollars died ten and a half hours early. Spot moved seven pips.
The structural half of yesterday’s piece was that none of the four expiry boards on the sheet expired after the event in its own currency, and the Australian board was the sharpest case: 4.969 billion Australian dollars, 2.1 billion of it at 0.6975, extinguished at 14:00 UTC with a quarterly inflation print due at 00:30 and a Reserve Bank of New Zealand decision at 01:00.
That is now history and it went exactly as described, in the dullest possible way. Spot was around 0.7021 before the cut. A reader at roughly 20:00 UTC has the Australian dollar 0.2 per cent higher at 70.28 US cents, and a later entry on the same page calls it “marginally higher” into the release. Seven pips. The 2.1 billion at 0.6975 expired about forty-six pips out of the money and the 1.2 billion at 0.7030 about nine pips out, so nearly five billion dollars of optionality was retired without a fight, and the quarterly inflation number it might have been hedging arrives ten and a half hours later into a book that no longer holds it.
Draw the operational conclusion rather than the market one. If your sizing rule reads “large expiries today, reduce,” yesterday was the day it was exactly backwards in the Australian dollar: the expiry was the thing that made today’s hours less cushioned, not more. The gamma that would have damped a CPI surprise was the gamma that expired first.
Two readers agree on the oil price to twelve cents and disagree on the day by half a point
A thing this desk keeps relearning in new instruments. One reader gives settlements of 99.90 dollars for West Texas Intermediate, down 1.86, and 113.88 for Brent, down 1.93 — minus 1.83 and minus 1.67 per cent. A second, a rolling snapshot taken at roughly 20:41 UTC, gives 99.78 and 113.84, down 2.0 and down 1.2 per cent. The prices agree to twelve and four cents. The Brent percentage move differs by about half a point, and working backwards the two imply prior closes of 115.81 and 115.22 — fifty-nine cents apart.
So the level is two-sourced and tight, and the change is not sourced at all until you know which prior close and which cut-off each reader used; one is a settlement and the other is a live snapshot, which accounts for part of it and we have not established how much. A price is not a move. We have published that sentence about an average without its window, a rate without its instrument and a consensus in the wrong field, and here it is in a commodity percentage.
It happens to dollar–yen on the same day and worse. One reader has 157.86 up 0.37 on the day, which implies a prior close of 157.49. Another, at 07:01 UTC, had 157.45 up 0.41 per cent, which implies 156.81. Sixty-eight pips of disagreement about yesterday’s starting point, between two publishers quoting the same pair. We are publishing the levels and declining to publish a day’s change for dollar–yen at all, because we cannot say which window either reader is differencing against. For the rest of the close: the Standard and Poor’s 500 at 7,722 up 0.60 per cent, the Nasdaq Composite 26,477 up 1.18, the Dow 52,378 up 0.03, the Russell 2000 2,818 up 1.16, with a second reader matching the three index percentages; the two-year note at 4.735 per cent up 2.6 basis points, the ten-year 4.992 roughly unchanged and the thirty-year 5.304; December gold 4,074.60 up 36.20 and silver 54.51.
What this slot structurally cannot see
Our own work order for this run asked for a second instance of the Tokyo fixing. It cannot be had, and the reason is a property of the schedule rather than of the sources. This note publishes at 08:00 Japan time. The Mitsubishi UFJ fixing is published around 09:55, and yesterday’s landed at 10:06. The Tokyo-morning slot fires roughly an hour and fifty minutes before the fixing exists, so a published fixing is a pre-London object and always will be. Yesterday’s 157.64 did come from the pre-London slot, and it is now confirmed from a second route: a different page of the same publisher carries a 27 October fixing of 157.64 as a related item.
That second route also set a trap worth describing, because we nearly walked into it. A search for today’s fixing returned two pages headlined with fixings of 157.96 and 157.77 — both entirely plausible levels for this week, both within thirty pips of spot. They are dated 1 October and 28 September. Four and five weeks old. Neither URL carries a date: the identifier is a monotone integer, so there is nothing in the address to read. Yesterday we wrote a rule telling ourselves to read the date segment of every dated slug. The version that would have caught this one is broader: a plausible level is not evidence of a current level, and when the address carries no date you have to open the page and find one before you use the number.
And the thing this slot is actually for. The Asian hours ahead of it hold three events in forty-five minutes — a Reserve Bank of Australia estimates hearing at 00:15 UTC, the September-quarter and September-monthly inflation prints together at 00:30, and the Reserve Bank of New Zealand at 01:00 — and then the Federal Reserve at 18:00 with a press conference at 18:30, and the Bank of Japan on Friday, expected to hold at 1.25 per cent. We were asked whether these Asian hours are thinner than the last two mornings. We cannot answer that and will not guess: nothing we read publishes a volume or spread series for the Tokyo session, and a range comparison across three days with different event loads is not a liquidity measurement. The measurement that would settle it is tick count or quoted spread in the hour before 00:15, and if your platform logs that, you have something no publisher we use does.
What this does not tell you
It does not tell you that expiry concentration pins spot. One session, one coincidence, and a moving average eleven pips away with an equally good claim on the same high. It does not tell you where spot was at the cut, which is the observation the whole scoring would need.
The expiry notionals remain single-sourced from one publisher’s board; a mirror confirmed the arrangement of the strikes and carried no sizes, for the third time. The 157.86 close, the index closes, the yields and the metals are one reader each unless stated, with the equity percentages matched by a second and the ten-year by a second. The oil prices are two-sourced and the oil percentage changes are, in effect, unsourced.
One figure we are deliberately not publishing as a fact. A reader quotes speculative positioning at 69,878 contracts net long yen, against the 103,023 this desk has leaned on since mid-September. If true that is a large unwind and it would matter. It carries no as-of date, and positioning data of this kind is measured at a Tuesday close and labelled by release date, so a figure without an as-of date is unusable for exactly the reason our own standing rule says it is — and the article it comes from contradicts itself elsewhere, describing a yield spread as narrowing by 0.147 points while its own table shows about 0.03. It is a lead, not a number. The positioning piece this desk has now deferred eight times needs a byline assigned in advance rather than another deferral, and the desks that own it are not publishing this slot.
Related
- Signals — range-versus-trend character in a forty-seven-pip session inside a fifty-pip strike cluster.
- EA presets — what an expiry-driven sizing rule does when the expiry precedes the event.
- FX event calendar — the 00:15, 00:30 and 01:00 UTC block and the 18:00 decision.
- Every expiry on the board dies before its own event — the claim this article marks.
- The fixing printed at 157.64, and so did Monday’s high — the coincidence we refused to build on, and the figure now second-routed.
- Two publishers named a pivot 46.6 pips apart — the same window problem in a level.
