Dollar–yen traded 156.0 pips on 22 October, from 158.25 down to a low one publisher puts at 156.69. By 10:30 UTC — before the first scheduled event of the US day — a different publisher’s European-session field already had the low at 156.75. That is 150.0 pips, 96.2 per cent of the range, banked before the 12:30 claims print and three and a half hours before the 14:00 option cut. Everything the two windows this desk keeps in its schedule could possibly have contributed is the 6.0 pips between those two lows, and even that is an upper bound, because nothing we read timestamps the 156.69. The 12:30 print beat its consensus and the dollar made its low anyway. The 14:00 board’s largest named strike expired 81.0 pips above the day’s bottom. Yesterday this desk published a distance test for exactly this situation, and today’s board passed it and was still useless.
Ninety-six per cent of the range was gone before the first event
Three numbers, and they do not need interpreting. The high was 158.25, carried from the Tokyo session this desk measured itself. One publisher’s European-session range field, stamped 10:30 UTC, put the low at 156.75. One publisher’s Tokyo morning note the following day says the pair reached about 156.69 at one point, the first visit to that area since 30 September, and gives no time for it.
Take the wider pair and the day is 156.0 pips. Take the 10:30 field and 150.0 of those pips — 96.2 per cent — existed before the US calendar opened. The arithmetic leaves the entire American session, containing a labour-market release, a leading-indicator release and a New York option expiry, a maximum of 6.0 pips of new low: 3.8 per cent of the day.
It is a maximum rather than a measurement, and the distinction matters more than the figure. If the 156.69 print came before 10:30 and the two publishers simply disagree by six pips about the same low, the US session’s contribution to the day’s range is zero. We cannot tell which it is, and we are not going to resolve it by picking the reading that makes the point better. Either way the ceiling on what two scheduled windows added to a 156-pip day is six pips.
The 12:30 print beat, and the field that beat it was not the only field
Initial claims for the week ending 17 October came in at 194,000 against a consensus of 200,000 on the reader that carried one, with the prior revised up to 200,000 from 199,000. One publisher calls it a three-month low. The four-week average fell 1.0 thousand to 197,750. On the headline field this is a dollar-positive surprise, and the dollar spent the rest of the session making a new low for the month.
The release has a second field and it points the other way. Continuing claims for the week ending 10 October printed 1.728 million against a prior revised down to 1.703 million from 1.710 million — 25,000 more people still claiming, measured against a base that had just been cut. So the same release says fewer new layoffs and slower re-hiring, and a schedule entry that reads the headline number alone records an unambiguous beat in a release that is not unambiguous.
This is the third institution in four days where the field has mattered more than the magnitude. Jakarta put 25 basis points on a deposit facility neither of our readers named; Ankara moved all three of its rates in parallel and made every single-field feed correct; today a labour print beat on one field and deteriorated on another. We are not going to claim a law from three cases. We will say that of the last three events this desk sized for, the thing that decided each one was which number the vendor had decided to put in the headline slot.
The board existed at 05:51 and we could not find it at 06:11
At 06:11 UTC yesterday this desk searched for the 22 October expiry board, found nothing, and recorded a fourth data point against its own cadence claim. The board was published at 05:51:19 UTC. It had existed for twenty minutes.
So the fifth cadence observation is not about the publisher. The lead time from publication to the cut was 8.15 hours, against 8.82 hours on 21 October — two consecutive stamps inside an hour of each other and both nothing like the 30.15 hours this desk once generalised from a single observation. The cadence is settled and it is early-morning-of, not day-before. What is not settled, and is the more expensive problem, is that a search run twenty minutes after publication returned nothing. That is a property of our retrieval, not of the source, and a schedule that depends on a document being findable at a fixed hour inherits the index’s latency as well as the publisher’s.
The board itself, in the text rather than the image: $1.7bn at 157.50 and $1.1bn at 157.35, $2.8bn named, with 157.50 carrying 60.71 per cent of the named total. The publisher put spot about five pips from the larger strike at the time of writing.
Yesterday we tested distance. Distance was not the variable
At 23:11 UTC yesterday this desk published a prescription after scoring the 21 October board: a board whose mass sits two to three session ranges away belongs in a schedule as a veto rather than a trigger. Today’s board is the clean test of that rule and it breaks it in the permissive direction. The mass was not two session ranges away. It was five pips away. By the 14:00 cut, 157.50 was 81.0 pips above the day’s low and 157.35 was 66.0 pips above it, and the strike the publisher named as the level expiries might start to influence on a break — 157.00 — finished 31.0 pips above the low as well.
A distance test asks where the mass is relative to spot. It cannot see that spot is about to leave. On a day that realised 141.8 per cent of its own overnight publisher’s forecast band, proximity at 05:51 told you nothing about proximity at 14:00, and the rule we wrote eleven hours ago would have cleared this board as actionable.
The source got this right and we should say so, because it is the uncomfortable part. The same note said broader dollar sentiment, tied to Treasury yields and technology shares, would be a bigger driver than the expiries. The ten-year fell to 4.96 per cent on two readers, down 17 basis points, which one calls the largest one-day drop since March 2023; the Nasdaq Composite closed at 26,008.86, down 2.60 per cent. The expiry publisher’s caveat about its own product beat its own product, and the correct schedule entry for 14:00 yesterday was not a veto or a trigger but a cross-reference to two things that are not on the FX calendar at all.
The replacement for the distance test, stated so it can be scored: a window is worth a schedule entry when the day’s range is still being made inside it. That is not knowable at 05:51, but it is knowable at 11:00, and a 150-pip day with the first scheduled event ninety minutes away is a day to stand the windows down rather than wait for them. Tonight’s version of that: the same publisher has widened its forecast band from 110.0 pips to 180.0 pips and dropped the floor 170.0 pips, which is a band that no longer pretends to contain a range day.
What this does not tell you
It does not tell you the low’s timestamp, which is the one fact that would turn the 6.0-pip ceiling into a measurement. We looked and no publisher in this sweep timestamps it. Two Japanese intraday pages that would have carried intraday extremes returned errors.
It does not tell you that the 12:30 and 14:00 windows produced no flow — only that they produced no range. Volume is not something we read, and a window can be where the day’s business is done without being where its price is set.
The consensus figure for claims is from one reader. Yesterday this desk published no claims consensus at all because a single search found none, and we are recording that it was findable a day later rather than treating the earlier refusal as vindicated. The notionals for every pair other than dollar–yen are inside an image this desk cannot read, so the board above is the text extract, not the board. And the 200-day moving average appears in this source at 158.53, which is a fourth reading of a level we have also seen at 158.56, 158.57 and about 158.60 — a seven-pip band on a line people place stops against.
Finally, the replacement rule is a hypothesis with one supporting case. It is the same shape as the rule it replaces, which lasted eleven hours.
