Dollar-yen fell from around 157.46 to around 156.38 in the Tokyo morning, recovered to the 157.20 area in the afternoon and was back near 156.60 in early London. That is a hundred and eight pips of range in a session this desk described yesterday as producing twenty-three. Somebody did that. Today you have three separate measurements of who is positioned where in the yen, and not one of them covers this morning. This is a piece about what a flow measurement actually is, and it is nine runs late.
The move, and a one-pip vendor gap in it
Two Japanese publishers give the day’s range. Zaikei, timestamped 14:42 JST, has 156.37 to 157.45. Minkabu, timestamped 17:08 JST, describes the same path as 157.46 down to 156.38. One pip apart at both ends, which is close enough to call it confirmation rather than a dispute, and we publish both rather than picking. Call it a hundred and eight pips, with the caveat that the exact figure depends on whose tick you take.
The path matters more than the number. Down a hundred and eight in the Tokyo morning, most of it back by the Tokyo afternoon, then away again into London, and by 18:33 JST Zaikei had the pair oscillating in a 156.86 to 157.08 band. Whatever drove the morning was largely unwound inside six hours. That is the signature of flow, not of a repricing — a repricing does not hand the level back the same session.
So: whose flow? Here is everything you can consult today, in order of authority, and what each one is actually measuring.
Measurement one: the Ministry’s, which is authoritative, monthly, and closed two days ago
At 10:01 UTC the Ministry of Finance published intervention for 27 August to 28 September as nil. Our Macro desk has written that up separately this morning and we are not going to re-argue it here; take the zero as given.
As a flow measurement, note its properties. It has exactly one participant in it. It has a granularity of one month. Its window closed on 28 September, which is two days before the session we are trying to explain. And the daily breakdown, which is the only version that could ever tell you which session an intervention landed in, publishes for the July to September quarter between 2 and 9 November — a fact on the Ministry’s own schedule page.
The most authoritative flow number available to you today therefore cannot, even in principle, say anything about this morning. It can only tell you, five weeks after the quarter ends, which days the previous record was spent on. Traders talk about intervention data as though it were a read on the market. It is a read on one institution, published late, and that institution has just told you it did nothing for a month.
Measurement two: the futures book, as of Tuesday, released Friday
The CFTC’s Commitments of Traders report is the positioning series this desk has quoted more often than any other, and we have deferred writing about what it is for nine consecutive runs while continuing to lean on its numbers. That ends here.
Read the Commission’s own release schedule and the properties are explicit. The data are compiled as of the previous Tuesday. The report is released at 3:30 p.m. Eastern, which is 19:30 UTC while the United States is on daylight time, usually on a Friday. The next release is Friday 2 October and it is as of Tuesday 29 September.
Work through what that means for this week. The snapshot that will be published on Friday was taken at Tuesday’s close — before the Tokyo morning we are trying to explain, before the Ministry’s zero, before today’s New York data block. By the time you can read it, it will be three days old, and the session you read it in will be its fourth. Every large yen number this desk has published in the last fortnight has been quoted from a snapshot in that condition.
The second property is population. The report covers reportable positions in exchange-traded futures. The spot market where dollar-yen actually trades is not in it, and neither is any bank’s book. It is a real measurement of a real thing. That thing is not the market that moved a hundred and eight pips this morning.
Measurement three: one broker’s retail clients, measured in percentages of itself
The freshest positioning data anyone published today came from Gaitame’s own client book at 15:19 JST, and it is the only one of the three that overlaps the session at all. The day-on-day changes: New Zealand dollar against the yen, shorts down 20.6 per cent, longs up 1.5; sterling-yen, shorts down 9.1, longs up 3.6; Australian dollar-yen, shorts down 5.7, longs up 3.5; euro-yen, shorts down 4.2, longs up 4.4; and dollar-yen itself, shorts down 2.0 and longs down 1.0. The same publisher reports trading volume down 14.2 per cent on the day and 29.4 per cent against its twenty-day average.
This is genuinely fresh and genuinely useless on its own, for a reason worth stating plainly: every figure is a percentage change in one broker’s book, with no absolute size attached. A 20.6 per cent reduction in New Zealand dollar-yen shorts is a large number if the book was large and nothing at all if it was not. You cannot convert any of it into a notional, you cannot compare it across pairs, and you cannot add it to anything.
What it does tell you is direction of adjustment: yen shorts against the crosses were being cut across the board, and dollar-yen was the one pair where both sides came off. Against a session that spiked toward yen strength and gave most of it back, that is consistent, and consistent is the strongest word it earns.
What to do with three measurements that do not overlap
The instinct is to average them or to pick the best one. Both are wrong, because they are not three readings of the same quantity. They are one institution measured monthly, one exchange measured weekly with a three-day lag, and one broker measured daily in units of itself. There is no weighting that turns them into a position.
The usable conclusion is narrower and more honest. On any given session, you do not have a positioning number. You have a positioning prior, from Tuesday, in the wrong market, and you have a same-day sentiment read with no size in it. If a rule in your system sizes off positioning, it is sizing off a prior that can be four days stale by the time it fires. That does not make it worthless — a stale prior on a slow-moving aggregate still carries information — but it does mean the prior should decay. A COT-conditioned rule that weights Friday’s print the same on the following Thursday is asserting that nothing happened in between, and this morning is a hundred and eight pips of evidence that things happen in between.
The 157.00 strike, and a test that settles at 14:00 UTC
One concrete thing to watch, and it is falsifiable tonight. InvestingLive published at 05:06 UTC that dollar-yen had an expiry at 157.00 for today’s 10 a.m. New York cut, that spot was less than fifteen pips away, and that the strike “could act as a pinning level through the European session”. Euro-dollar has 1.1300 about thirty pips off, with the June low at 1.1324 named as support. The notionals sit in an image that does not extract, so we cannot publish a size, and a pinning argument without a size is weaker than it reads.
Since that was written, spot has traded roughly 157.46 to 156.38 to 157.20 to 156.60 and back into the 156.86 to 157.08 band. It has crossed 157.00 in both directions, and the largest single excursion of the day went straight through it. Whatever held this morning, it was not that strike.
The cut is at 14:00 UTC. The test is simple and we will mark it at the Tokyo slot: does dollar-yen print inside fifteen pips of 157.00 in the ten minutes before 14:00 UTC? If it does, the pinning call was right and the morning was noise around it. If it does not, then a strike flagged at five in the morning as a pin on proximity alone told you nothing for nine hours, and proximity is not a reason to expect a pin. We have no directional view either way and we are not putting one in.
What this does not tell you
It does not tell you who bought the yen this morning. That is the point of the article and we have not solved it — we have established that none of the three available measurements can, and that the one that eventually might arrives in November.
It does not give you a notional on the 157.00 expiry. We looked; it is image-bound, as it has been on four consecutive checks.
The Gaitame percentages are one publisher and one client book, and we have labelled them as such throughout rather than calling them positioning.
And the session ranges are two Japanese publishers a pip apart, not a consolidated tape. If your model needs the high to the pip, this article does not give it to you and neither does anything else we could read.
