You have thirty minutes until five American releases land in the same minute, two hours until the New York cut, and three separate measurements of how the market is positioned. One of them was taken eight days before the inflation print that repriced the Federal Reserve. One of them exists only inside a picture. The third is a column of percentages with no denominator published anywhere on the page. All three are current, all three are the best available, and not one of them can tell you how big anybody’s book is.
The futures positioning data predates the print that matters
The most recent Commitments of Traders report on Japanese yen futures is measured at the Tuesday close of 6 October. Two independent readers give it identically, to the contract: non-commercial longs 190,368, shorts 128,043, a net long of 62,325 contracts, a week-on-week net change of 6,885, open interest 368,138. That is a long/short ratio of 1.49 and a net position equal to 16.9 per cent of open interest. At 12.5 million yen a contract it is about 779 billion yen of notional, or roughly 4.9 billion dollars at this morning’s spot.
We have been promising this baseline for four slots and it is now set, second-sourced, with no disagreement between the two readers. Here is what it is worth.
It was measured eight days and sixteen hours ago. The September consumer price index printed on 14 October at 12:30 UTC and doubled the market-implied probability of an October Federal Reserve increase. The COT snapshot predates that print by seven days and sixteen hours. It cannot contain the reaction, because the reaction had not happened.
The next report lands tomorrow at 19:30 UTC. It is measured at the 13 October close — sixteen and a half hours before the print. It cannot contain the reaction either. The first positioning snapshot that includes what traders did after the inflation data is measured at the 20 October close and published on 23 October, eight days and seven hours from now.
So the honest statement about speculative yen positioning today is this: the market was net long 62,325 contracts of yen before the event that everybody is trading, and nobody will publish what it did next for another eight days.
The two readers agree on the numbers and disagree about when they became public
Both of our sources carry the same figures. They do not carry the same account of the report’s own timing.
One states that the data is as of 6 October 2026 and was “released on that date at 16:30 EST”. That is wrong three times over. The Commitments of Traders report for the 6 October close was published on Friday 9 October, not on the Tuesday it measures. The release hour is 15:30 Eastern, not 16:30. And there is no Eastern Standard Time in October — the United States is on daylight time until 1 November.
The other reader states the rule correctly — every Friday, data from the previous Tuesday — and then gives no release date for this particular report at all. A third page we read describes the schedule accurately and carries no figures whatsoever.
This is the measurement-labelling failure this desk has written about four times in two weeks, and it is now on the positioning series. The values are fine. What is wrong is every surface’s account of when the values came into existence, which is the only thing that tells you whether they are usable. A vendor that collapses the measurement date into the release date has quietly erased three days of staleness from your risk picture.
We finally got the strikes. The sizes are still inside a photograph.
For seven consecutive slots this desk has reported that it could not publish an option expiry ladder. This morning the page loaded. At 05:56 UTC it named a dollar–yen strike at 158.00, about five pips from spot at the time; a euro–dollar strike at 1.1250 described as chunky, roughly forty pips away; and a euro–sterling strike at 0.8480, about eleven pips above spot.
The notional amounts are in an image file. The text of the page carries none of them. A mirror we checked previously does not reproduce the table.
A strike without a notional tells you where the gravity is and nothing about its mass. Two hours from now, at the 14:00 UTC cut, a 158.00 expiry could be two hundred million dollars that nobody notices or two billion that pins spot through the American morning, and the published record this morning does not distinguish those cases. The 12:30 release block lands ninety minutes before the cut, which is precisely the window in which the size would matter most.
Eight slots. We will keep reporting the strikes we can read and keep saying plainly that the sizes are unavailable, because the alternative — implying a pin we cannot measure — is worse than admitting the gap.
The retail book moved hard and the page will not say how much
A Japanese broker published its retail positioning snapshot at 06:19 UTC, covering the previous European session through the Tokyo day. Across all pairs the book is 56 per cent long and 44 per cent short, with sell positions down 3.8 and buy positions up 0.6 against the previous day. Dollar–yen accounted for 68.7 of turnover, a reading down 10.8 on the day.
The per-pair changes are where the story is, and they are not small. Short positions in New Zealand dollar–yen fell 33.9. Euro–yen shorts fell 17.9. Sterling–yen shorts fell 16.3. Mexican peso–yen 9.0, South African rand–yen 6.9, Australian dollar–yen 6.8. The average cut across those six cross-yen shorts is 15.1.
Dollar–yen shorts fell 1.8, and longs fell 3.5. The crosses were dismantled by an order of magnitude more than the dollar pair — the New Zealand figure is nearly nineteen times the dollar–yen one.
Now the problem. Every one of those numbers is a percentage change. The page publishes no absolute position sizes, so a 33.9 reduction in New Zealand dollar–yen shorts is unmeasurable in money. It is also not labelled: we cannot tell from the page whether the 68.7 turnover share moved by 10.8 percentage points or by 10.8 per cent of itself, and those are different claims about the same number.
Against that, the one figure on the page that does have a visible denominator says the opposite thing. Total trading value was up 5.2 against the previous day and down 20.1 against its own twenty-day average. A book being torn apart across six crosses and a turnover figure a fifth below its own recent norm cannot both be the dominant description of this session, and you cannot reconcile them, because the sizes are not published.
What to actually do with three broken rulers
Not nothing, and not what the numbers superficially invite.
First: do not use COT as a crowding input this week. It is not stale in the ordinary sense of being a few days old. It is stale in the specific sense that it was taken before the event your position is about, and so is the one arriving tomorrow. If your system has a positioning filter, it is running on pre-print data until 23 October, and it should either be told that or switched off rather than left to imply a crowding level it cannot see.
Second: treat the 158.00 strike as a location, not a force. Mark it, do not size to it. A level you have a strike for and no notional for belongs in the same category as a moving average whose convention nobody states — worth watching, not worth leaning on.
Third: if you use retail positioning as a contrarian signal, note that today it is telling you about direction of change and not about magnitude, and that the only magnitude on the page points the other way. A signal with no denominator cannot be scaled, and a signal that cannot be scaled cannot size a trade.
Fourth, and this is the part that costs something: the next ninety minutes contain five releases and then a cut. It is the one window today where positioning data would actually earn its keep, and it is the window in which all three of our measurements are least usable. The correct response to a measurement you cannot trust is a smaller position, not a cleverer inference.
What this does not tell you
It does not tell you which way dollar–yen goes after 12:30. We have no view and are not offering one. The pair set a weekly low at 157.58 in Asian hours and was back at 158.00 by the early European session, and we cannot decompose that into flow, yields or the data ahead.
The COT figure is a futures measurement on one exchange. It is not the yen position of the world, it does not include the over-the-counter market where most of the volume lives, and a net long of 62,325 contracts tells you nothing about leverage or holding period.
The retail positioning data covers one broker’s platform. It is not the Japanese retail market and it is certainly not the market. We have published its changes because they are large and consistent across six pairs, not because we can establish they are representative.
And we did not read the option expiry notionals. We are not estimating them, we are not inferring them from the description of one strike as chunky, and if the 158.00 expiry turns out to have mattered this afternoon, we will not claim to have told you so.
Related
- Event calendar — the 12:30 block and the 14:00 cut, in UTC.
- Signals — what we act on when the positioning inputs are unusable.
- Presets — filters and exposure settings for clustered release windows.
Earlier from this desk: a claim we made about an issuer that turned out to be about our own reader, and a session that traded ninety pips against sixty-one priced.
