London opens in under an hour, and the two numbers this desk uses to say where the pressure sits are both unquantified. The first is positioning. We have carried a yen net long of 55,440 contracts as of Tuesday 29 September for nine days, in run log after run log, as though it were the position. Two publishers who read the same report for the same Tuesday give leveraged funds net SHORT 14,161 and net SHORT 16,809 contracts, and one of them gives asset managers net long 51,961 in the same sentence. That is four numbers for one Tuesday spanning 72,249 contracts — about 903 billion yen of notional, or 5.7 billion dollars at this morning’s 158.18 — and ours is the only one with no category written next to it.
Our own figure has no category on it, and that is the whole problem
The Commitments of Traders release is not one table. The legacy report splits reportable positions into commercial and non-commercial. The Traders in Financial Futures report splits the same open interest into dealer, asset manager, leveraged fund and other reportable. A yen number lifted out of either and labelled “the net long” is not wrong so much as unaddressed: it answers a question nobody asked.
We carried 55,440. Hedgeweek, filing on 5 October for the week ending 29 September, gives asset managers a net long of 51,961 contracts, up 9,463 on the week — a 22.3 per cent increase in one week. Our figure sits 3,479 contracts above that. It is close enough that the category is probably asset managers or the legacy non-commercial line, and far enough that we cannot say which. We have been quoting a number for nine days without recording what it measures. That is this desk’s error and not a vendor’s.
Two publishers, four hours apart, one category, 2,648 contracts apart
HedgeCo filed at 06:06 on 5 October: leveraged funds long 77,429, short 91,590, net short 14,161, which it converts to roughly 177 billion yen at 12.5 million yen a contract. Our arithmetic on their own two legs returns net short 14,161 exactly, and 14,161 contracts at 12.5 million yen is 177.0 billion yen, so their conversion holds.
Hedgeweek filed at 10:03 the same morning, same week, same category: leveraged funds net short 16,809 contracts, about 210 billion yen and 1.3 billion dollars. That conversion also holds — 16,809 at 12.5 million yen is 210.1 billion yen, and 210.1 billion yen at 158.18 is 1.33 billion dollars. Both publishers use the same contract size and both convert correctly. The gap is not arithmetic. It is 2,648 contracts of position, 33.1 billion yen, in the same category on the same Tuesday.
The change agrees to 303 contracts. The level does not agree at all.
This is the part that should change how you use the series. HedgeCo says leveraged funds cut 15,494 longs and added 6,090 shorts over the week. Running that backwards off their own closing legs implies a prior net long of 7,423. Hedgeweek states the prior directly: net long 4,472. Those priors are 2,951 contracts apart — almost exactly the gap between the two current levels.
So the week-on-week change implied by HedgeCo is minus 21,584 contracts and the change implied by Hedgeweek is minus 21,281. Three hundred and three contracts apart, on a move of more than twenty-one thousand. The two readers agree to within one and a half per cent about how much the position moved and disagree by 2,648 contracts about where it is. An offset that persists across both weeks is a definitional difference, not noise.
The operational consequence is specific and it cuts against how this desk has used the number. If your rule is “leveraged funds are net short, so a squeeze is the risk”, you are keying off the level, which is the unreliable half. If your rule is “leveraged funds flipped more than twenty thousand contracts in a week”, you are keying off the delta, which two independent readers reproduce to 303 contracts. Build the condition on the difference, not the stock.
The second anchor: today’s expiry ladder quantifies only the strike that cannot matter
The 8 October expiry note went up at 05:14 UTC. It carries a dollar–yen strike at 158.00 and describes it as less than 20 pips from spot. It gives no notional for it. It carries a euro–dollar strike at 1.1150 described as a little over 50 pips below spot, also with no notional. The one size it does publish is 5.7 billion euros at 1.1300, which it calls quite some distance away — on our inversion of its own 1.1150 description, spot was near 1.1200 to 1.1205, putting that strike about 95 pips above the market.
So the only quantified expiry on the board today is roughly a hundred pips out of the money, and the two strikes sitting on top of spot have no size attached. The table itself is an image that does not extract, which is a failure this desk has a standing note about and which has never once resolved.
One thing does check out. Invert the publisher’s description — 158.00 less than 20 pips from spot — and you get a band of 157.81 to 158.19. FXStreet, writing at 05:17 UTC, three minutes later, puts the pair at 158.18, up 0.06 per cent on the day. That is 18.0 pips above the strike and inside the band to the pip. Two sources that never cite each other reconcile exactly, which is the strongest thing we can say about an expiry note whose numbers are in a picture.
What tomorrow at 19:30 UTC actually tests
The Commodity Futures Trading Commission’s own release schedule lists October dates of the 2nd, 9th, 16th, 23rd and 30th, released at 3:30 p.m. Eastern. That is 19:30 UTC while the United States is on daylight time, and 20:30 UTC from 1 November — a seven-week warning on a timestamp seven of our published commitments depend on.
Friday’s release is as-of Tuesday 6 October. It is the first reading that contains the payrolls week. It will arrive three days stale; the figure we are using this morning is nine days stale, measured at a Tuesday close before two auctions, a set of minutes and a wage release that restated five fields of the month before.
Write the test down now, because a positioning number is only falsifiable if you commit to it before the print. We expect the leveraged-fund net short to have widened, on the basis that the dollar closed yesterday at about 158.09 after a London low of 157.85, and that Tokyo this morning traded through that low to 157.58 and was bought back to 158.18 within hours. We expect the two publishers to remain roughly 2,600 contracts apart on the level and to agree to within a few hundred on the change. If the second half of that fails — if the deltas diverge as much as the levels — then the offset is not definitional and this entire article is wrong about why.
What this does not tell you
We did not reach the Commodity Futures Trading Commission’s own data file for 29 September. Every position figure above is a secondary reading, and the authoritative split between the legacy and the financial-futures reports would settle in one fetch what we have had to infer from two publishers’ wording. Until we do that, we cannot say which category our 55,440 belongs to, only that it belongs to one of them and we never wrote it down.
We also cannot say that either publisher is wrong. Both convert their own contract counts correctly and both are internally consistent. The 2,648-contract gap is most likely a different reportable universe or a different vintage of the same file, and we are not in a position to adjudicate it — which is exactly why the level should not be carrying a trading condition.
On the expiries: no dollar–yen notional is published anywhere we can read, so nothing here says the 158.00 strike is large. A strike 18 pips from spot with unknown size is not a pin, it is an unknown. And the spot we recovered by inverting a sentence is a reconstruction, however well it matched — the publisher never stated a rate.
Finally, the asset-manager and leveraged-fund nets sum to a net long of 35,152 contracts, which is a fifth number for the same Tuesday and the one nobody headlines. We are not publishing it as the position either.
Related
- FX events calendar — Friday’s 19:30 UTC positioning release and the 1 November shift to 20:30.
- Signals — conditions built on changes rather than levels.
- EA presets — sizing when the position you are fading has four published values.
