The Commodity Futures Trading Commission’s first positioning snapshot containing the Bank of Japan’s rate rise landed on Friday evening, and it says speculators cut their net long in the yen by 48,377 contracts in a single week — from 120,359 to 71,982, a reduction of 40.2 percent. That is the headline every desk will quote this week. It is also the least interesting number in the report. The interesting one is the split: of that 48,377-contract swing, 45,677 contracts were longs getting out and 2,700 were shorts getting in. Ninety-four percent of the move was people leaving, not people arriving. Those are different trades, they behave differently on the next shock, and the word “net” hides the difference completely.
The number, read at the source rather than off a summary
The Commitments of Traders report for positions as of Tuesday 22 September gives Japanese yen non-commercial longs at 192,274 contracts and shorts at 120,292, with 4,376 in spreads, against total open interest of 378,701. Subtract and the net long is 71,982. The report’s own change line, measured against 15 September, reads longs minus 45,677 and shorts plus 2,700 — which implies a prior net of 120,359.
We did the subtraction ourselves rather than take a vendor’s figure, and we recommend you do too, because this is the week to check. Two readers published numbers for this release and both of them are exactly right. One described speculators as having reduced exposure “markedly to nearly 72.0K contracts”; our arithmetic on the Commission’s own table gives 71,982. Another vendor’s release history carries 120.4K as the prior; the Commission’s change line implies 120,359. Both round correctly. This desk has spent the better part of three weeks publishing vendor disagreements — a 40 basis point split on Canadian inflation, three feeds giving three probabilities for one binary event, four vendors carrying four different histories of the Bank of Japan’s core measures, a market wrap contradicting its own ticker. So it is worth saying plainly when the opposite happens. On this release, checked against the issuing agency, the vendors are right to the contract.
Ninety-four percent of the cut was an exit, not a new position
Here is why the composition matters more than the net. A long that liquidates has already taken its decision; the contract is gone and the holder has no further business with the trade. A short that opens has a decision still in front of it, at a level, with a stop. Two books that both show minus 48,377 on the net line can be positioned in completely opposite ways for what happens next, and the one dominated by exits is the one with less fuel in it.
Apply that to the week just ended. If the 48,377 had been mostly new shorts, Friday’s 202-pip fall in dollar-yen — 158.95 down to 156.93 — would have been running into a wall of profitable positions looking for a place to cover, and the bounce off the low would have been violent. It was not violent. Spot has recovered 47 pips of a 202-pip range by the Tokyo open, which is 23 percent, and it took a weekend to do it. That is consistent with a book that emptied rather than reversed.
We are stating that as consistency, not as proof. The COT tells you about futures on one exchange; most of the yen risk in the world is not there.
The snapshot is from Tuesday, and the week’s two biggest yen events came after it
This is the part that gets forgotten every single time positioning gets quoted, so we will put it in a sentence you can copy. The data is measured at the close of business on Tuesday 22 September. It was published on Friday 25 September. Between those two points: Japan’s finance minister disclosed at roughly 02:16 UTC on Friday that the US president had raised yen weakness with the prime minister; dollar-yen printed 158.95 and then 156.93; Iran put a seven-day plan to reopen the Strait of Hormuz on the table and crude fell; and the yen had what one wire called its best day in two weeks.
None of that is in the number. A desk that reads “net long cut 40 percent” this week and sizes against it is sizing against a book that existed three days before the two events that moved the currency. The honest use of this release is as a starting condition for Friday, not as a description of where anyone is now.
The release time, settled: 19:30 UTC, not 20:30
On Friday this desk published that it could not state when the COT is released, because one widely used calendar had it at 20:30 UTC while a separate finding of ours — a publisher rendering Eastern Standard Time and labelling the column “ET” in September — implied 19:30. We refused to print either and said the Commission’s own schedule would settle it. It does. The Commission states that the report “is generally published each Friday at 3:30 pm Eastern Time (US), using the data from the immediately preceding Tuesday of that week.”
Three thirty pm Eastern in late September is Eastern Daylight Time, so the release is at 19:30 UTC. The calendar carrying 20:30 is an hour late. The preview page carrying “2:30 PM ET” is an hour early and is early for exactly the reason we said — it is rendering standard time under a daylight-time label. Both vendors are wrong, in opposite directions, by the same hour, and the primary source costs one fetch. If you have a news filter with a COT window in it, it is in the wrong place.
And “the yen net position” is not one number, because it is not one report
One more thing to put in your notes before you quote a positioning figure at anyone. The Commission published a second report for the same Tuesday, on the same contract, under a different trader taxonomy. In the Traders in Financial Futures breakdown, yen Dealer Intermediary positions are long 56,693 and short 164,977 — a net short of 108,284 contracts, against the same 378,701 open interest.
Plus 71,982 and minus 108,284, same instrument, same date, both from the issuing agency, neither wrong. They are counting different people. This is not a vendor failure and there is nothing to reconcile; it is simply that “speculative positioning in the yen” is an incomplete sentence until it names a report and a category. We have watched this desk argue about whose consensus is right for three weeks. The COT version of that argument is avoidable, because the ambiguity is in the question.
Marking our own forward test — and it is void, not lost
On Friday morning this desk set a falsifiable claim in print. We said dollar-yen’s 45-pip fade after the Trump disclosure meant the session did not believe it, and that where the pair closed London and New York would settle whether that reading was right.
It closed lower. Spot went a further 67 pips below the 157.60 we last marked, to 156.93, and Monday’s Tokyo open finds it at 157.06 to 157.37 on one reader and 157.39 on another. On direction, the disbelief reading lost.
Except the pair did not fall because the market changed its mind about the disclosure. It fell because Iran offered to reopen the Strait of Hormuz, crude dropped, and the long-end inflation trade that has been carrying the dollar came off with it. Our test asked whether the market believed a piece of Japanese political news and then measured a price that moved on Iranian news. So the honest mark is not that we were wrong. It is that the test was badly built: it specified an outcome variable that three different causes could produce, and it did not specify a control. A test you cannot read is worse than a test you fail, because a failure teaches you something. We are recording this one as void and we are not claiming the direction as a result.
What this does not tell you
It does not tell you where anyone is positioned today. The snapshot is six days old as you read this and predates the largest yen move of the month.
It does not tell you that 71,982 is a small position. Against its own history it is not: one reader puts the exposure near the 85th percentile and the net near the 88th, and a 40 percent cut from a crowded long can leave you with a still-crowded long. We have not verified those percentile figures independently and they are that publisher’s calculation, not ours.
The long-versus-short composition argument above is an inference about behaviour, not a measurement of it. We can show you that 94 percent of the net change was long liquidation; we cannot show you what those holders intend next, and a book that emptied can refill in a session.
We have not read the CME FedWatch tool directly, and the October Federal Reserve probability this desk has been quoting at around 67.5 percent still reaches us at two removes. That remains open.
Nothing here is a view on direction. Positioning data is a sizing input. It has never once told this desk which way to go.
Related
- FX events calendar — the week’s releases, in UTC, including Wednesday’s Ministry of Finance intervention data at 10:00 and Friday’s US payrolls at 12:30.
- Signals — how we frame entries around scheduled release windows.
- Expert advisor presets — filter and exposure settings for news windows.
