Two Level Sets, One Cross, Sixty-Seven Minutes, Nothing in Common

Sixty-seven minutes, one cross, one publisher, zero shared levels

We counted. The 04:56 piece publishes seven price levels: 178.20, 180.00, 180.75, 182.90, 175.65, 175.34 and 172.27, each attached to a named construction — a Bollinger middle band, a psychological round number, an upper band, a 100-day moving average, a lower band and two dated 2025 lows. The 06:03 piece publishes two: 176.66 as a target and 177.66 as the level whose breach kills the view. The intersection of those two sets is empty.

That alone is unremarkable — different methods give different numbers. What matters operationally is the width. Spot sat near 177.30. The 04:56 set puts the nearest level above at 178.20, ninety pips away, and the nearest below at 175.65, one hundred and sixty-five away: a corridor two hundred and fifty-five pips wide. The 06:03 set puts the boundaries thirty-six pips above and sixty-four below: a corridor one hundred wide. One is 2.55 times the other. A stop placed outside the corridor, a position sized to it, or a filter that suppresses entries inside it will behave completely differently depending on which of two same-morning articles your process happened to read.

Neither piece mentions the other. Neither is wrong. The fault is in treating “the levels” as a property of the market rather than a property of the article.

The expiry pin rests on a spot price two other readers do not recognise

At 01:23 UTC an order-flow page published today’s New York cut expiries and named two euro–dollar strikes, 1.1200 and 1.1225. Its argument is that the pair of strikes sandwiches spot, sitting within fifteen pips on either side, and may therefore contain price into the cut. Twenty-five pips separate the strikes, so “fifteen either side” puts the author’s spot somewhere around 1.1213.

Six minutes earlier, at 01:17 GMT, a second publisher put euro–dollar at around 1.1240. Four minutes later, at 01:27, a Tokyo desk note gave the pair peaking at 1.1221 and softening to about 1.1211. Three published accounts of the same price inside ten minutes: roughly 1.1213 implied, 1.1240 stated, and a 1.1210 to 1.1221 range. The outlier is nineteen pips above the Tokyo peak and thirty above its floor, and twenty-seven and a half pips above the strike midpoint the pin argument needs.

By 05:27 UTC the Tokyo session range was given as 1.1208 to 1.1241. That range contains both strikes. The upper strike sits sixteen pips below the session high and the lower strike eight pips below the session low, which is to say the band was traded through at the top and never tested at the bottom, four and a half hours before the cut it was supposed to govern. We are not saying the pin failed — a ten-o’clock New York cut is not scored at half past five in the morning. We are saying the premise was contested at publication, and nobody reading the expiry page alone would know that.

The notionals, again and as on every observation this channel has made of this page, are inside an image. There is no quantity on the page. A strike without a size is a location, and we have said so often enough that it is now a house position rather than a complaint.

The calendar carrying today’s European data has no time column at all

The week-ahead calendar we read for this slot lists five releases for Friday: euro-area balance of trade for August, and then the September final inflation set — core annual, headline monthly, headline annual — plus American industrial production. It gives a previous and a consensus for every one of them. It gives a release time for none of them, because the table has no time column. Its header declares the whole thing is rendered in UTC plus seven.

So the document tells you what timezone it is in and then publishes nothing that needs one. The issuing agency is no better on this specific field: the euro-area flash for September, published on 2 October, states that the full September data follows on 16 October and states no hour. We have now recorded this failure mode nine times in our own archive, across separate publishers and more than one central bank, and the pattern is stable: the consensus field is almost always populated and the time field almost never is. If your news-window filter is built from a vendor calendar, it is being built from the one column those vendors do not reliably fill.

For the record, the flash the final is being measured against: headline 3.8 per cent annual in September against 3.2 in August, with energy at 18.8 from 14.3, services 3.2 from 3.0, food alcohol and tobacco 1.4 from 1.1, and non-energy industrial goods 1.1 from 1.2. A final revision to a flash is not usually tradeable. A six-tenths jump driven by a four and a half point move in one component is worth knowing before London prices it either way.

The 200-day is on its fifth published value this week

A Tokyo note stamped 08:07 JST gives the dollar–yen 200-day at about 158.58. That is the fifth reading this desk has logged inside a week: 158.50, 158.56, 158.57 and now 158.58, with two of them from the same publisher seven minutes apart yesterday. An eight-pip band.

Spot in the Tokyo afternoon sat in the 157.80s, with the session given as 157.79 to 158.18 by one reader and 157.93 to 158.17 by another — the highs one pip apart, the lows fourteen. Against the band, spot is sixty-five to seventy-three pips below. We keep publishing the band rather than a number because the direction of the answer does not change anywhere inside it, and that is the only reason the ambiguity is tolerable. It would not be tolerable for a rule that triggers on a cross of the average, because such a rule would fire on eight different pips depending on which publisher fed it.

What to actually do in the fifty minutes before the open

Not a directional instruction — a plumbing one. Three of the inputs a pre-London process typically consumes have failed this morning in three different ways, and all three failures are in the field rather than in the market.

First, if a level enters your system from a text feed, record which article it came from and at what hour, because this morning the same publisher served two incompatible sets on one cross inside seventy minutes. Second, if an expiry informs your sizing, record the spot the author was looking at, because today the pinning inference depends on a price nineteen to thirty pips away from what two other readers published in the same ten minutes. Third, if a news window comes from a vendor calendar, the window is the field most likely to be missing or wrong, and today it is simply absent from a table that nonetheless declares its timezone.

The cheap defence for all three is the same: store the publisher and the timestamp next to every number, and treat any figure that arrives without both as a characterisation rather than a measurement. Then test whether your rule survives the worst case across the readings you hold — the worst eight pips on the 200-day, the worst thirty on spot, the full absence of a release time.

What this does not tell you

It does not tell you where euro–yen or dollar–yen is going, and we are not offering a view on either. It does not tell you that any of the three publishers named here is wrong; each number may be a correct reading of a differently-constructed field, and in the euro–dollar case we cannot rule out that the 1.1240 figure is a stale carry of the previous day’s high, which one source puts at exactly that level — we could not establish it either way and have not treated it as settled.

The expiry claim is not scored here and will not be scoreable until well after the 14:00 UTC cut, and even then only against a spot print that this channel’s sources have repeatedly published more than one way. We have no notional for either strike and therefore no measure of how much pressure, if any, sits at them. We also have not read the option board or any order book directly at any point this week; everything in section two is a reading of readings.

The euro-area final inflation figure had not been released when this was written, and the American industrial production number lands at 13:15 UTC, seven hours out. The Bank of Japan governor’s speech, read on his behalf by the deputy governor, falls at 06:30 UTC — about twenty minutes after this went to the desk and before London opens. Nothing here anticipates it.

Related

• FX events calendar — release windows, in UTC, with the hour attached.
• Signals — how we frame entries when the level set itself is contested.
• EA presets — filter and news-window configuration.
• The low is unanimous at 157.55; the close is published two ways — this morning’s session piece, and the source of the 200-day band.
• Three positioning numbers, all broken differently — where the image-locked notional problem was last logged.