At 06:18 UTC this morning this desk published that dollar–yen had one named support below spot — an Ichimoku cloud lower bound at 156.06 — and that it had not been tested, because Tokyo’s low was 156.40, a comfortable 34.0 pips above it. Everything that followed in that article, including a 72.0-pip void to the next named level, was framed as a statement about a test that had not happened. The test had already happened. Four publishers now put the session low at 156.08, which is 2.0 pips above the support, and it printed in the Tokyo morning — roughly five hours before we called it untested. The 156.40 we used came from a desk note timestamped 00:04 UTC, and the pair took another 32.0 pips off the low between that note and 01:00 UTC. We published a stale range field as a live one, and the number we got wrong was the only number in the article that mattered.
The level held, and it held by two pips
Start with what the market did, because it is the better half of the story. The Tokyo-session low is 156.08 on four independent readers: one Japanese broker’s 10:00 JST market summary, which also gives a reference high of 156.82 and spot at 156.48 against a New York close of 156.77; the same broker’s 10:44 JST outlook, which calls 156.08 the lowest print since the 18th of last month; a second publisher’s 16:40 JST technical note; and a third reader’s mid-morning outlook. None of the four gives a timestamp for the low itself, which is the one thing we would still like. All four agree on the figure.
Against a cloud lower bound at 156.06, that is a 2.0-pip miss. We will not dress it up as a successful prediction, because we did not predict it — we said the level was 34.0 pips away and untested. But as a measurement of the level itself it is about as clean a result as this desk has recorded: a support named by one publisher at 10:58 JST yesterday stopped a decline two pips short, and the pair then recovered 108.0 pips to a London-session high of 157.16.
What the hold does not do is settle the 72.0-pip void beneath. 156.06 to the 17 September low at 155.34 is still 72.0 pips of unnamed space, and it is still a claim about what happens after a break that has not occurred. Approaching a level to within two pips and bouncing is evidence that the level is real. It is not evidence about the shape of the ground below it.
How we published a low that was 32 pips too high
The mechanism is worth spelling out because it is cheap to repeat. The pre-London sweep read a Japanese desk note timestamped 09:04 JST — 00:04 UTC — which gave the Tokyo high at 156.81 and the low at 156.40. That was correct when it was written. By 10:00 JST a different publisher’s summary had the low at 156.08 and the high at 156.82, one pip above the earlier high. So the session took 32.0 more pips out of the low in the 56 minutes between the two notes, and our 06:18 article, written five hours after that, used the earlier figure.
The error in the published distance is therefore 32.0 pips on a claimed 34.0 — the figure we printed was 17.0 times the real one. That is not a rounding problem or a vendor disagreement. It is a session high-low pair treated as a fact about the session rather than as a reading taken at a time.
This morning’s other article made exactly this argument about levels. It retracted a 41.0-pip cloud-bound “dispute” on the grounds that an Ichimoku leading span is recomputed every session, so two readings a day apart are two different objects sharing a name, and it concluded that every level this desk carries needs an as-of date. The conclusion was right and it was too narrow. An intraday high or low is the same kind of object: it is the extreme of a window that is still open, and a publisher’s range field is a statement about the window up to the moment of publication and nothing after it. We applied the rule to moving averages and missed it on the one field in the article that was still moving.
The rolling-window point, now with a number on it
While we are here: the same two publishers give the 21-day moving average at 157.88 today. Yesterday’s pre-London article put it at 158.02. That is a 14.0-pip move in a named level in twenty-four hours, from the same indicator on the same instrument, and it is the cleanest quantification this desk has of why a level carried without a date decays. Fourteen pips is not noise when the whole argument about where spot sits relative to resistance is measured in tens of pips.
The 200-day is worse, and it is our own long-running item. This desk has carried a 7.0-pip band for that level — 158.53, 158.56, 158.57 and about 158.60 — and published it as a measurement disagreement between vendors. Today adds a fifth value, 158.55, from the publisher whose 16:40 JST note we used above. We are not going to call that a vendor dispute either. A 200-day moving average is a function of a 200-session window, so a reading from Monday and a reading from Friday are different numbers by construction, and until somebody collects all five readings with their dates attached the honest position is that we do not know how much of the 7.0-pip band is vendor spread and how much is date spread. It is on the list to resolve, with dates this time.
What actually stopped the decline, and what is holding it up now
Two publishers give the same reason for the low, and it is not a level. Both say the decline paused when dollar buying came in around the Tokyo fixing — real-money demand on a gotobi date, the fifth-and-tenth settlement convention that concentrates Japanese corporate flow at the 10:00 JST rate. One of them names it explicitly as the reason the pair stabilised after 156.08. If that is right, the cloud bound did not hold the market; a settlement window did, and it happened to sit two pips below where the cloud bound was drawn. We cannot separate the two from anything we read, and anyone treating 156.06 as a tested, defended level should know that its defence has a plausible alternative author.
The background was risk-off. The Nikkei fell more than 1,400 points at one stage before recovering over 1,200 from its low, which is a 1,400-point swing in an index and a proximate cause for yen buying that has nothing to do with Japanese rates. By early London the pair had printed 157.16, 20.0 pips short of the 157.36 cap two publishers have now named twice. One reader puts spot at 156.64 in the early US morning: 56.0 pips above the low, 52.0 pips off the London high, and 13.0 pips below yesterday’s 156.77 close. The 156.80 that acted as support from mid-Tokyo is now 16.0 pips above spot, which makes it the first thing to watch into the 13:45 UTC American data.
The sizing consequence, which is the only part of this that is actionable: the first named resistance above spot is 157.36 at 72.0 pips, and the 21-day sits 124.0 pips away. The first named support is 156.80, 16.0 pips below, and then 156.06 at 58.0 pips. For the first time this week the nearest level in either direction is close enough to be a stop rather than a target, and it is on the downside.
What this does not tell you
Nothing here is a view on direction. The 156.08 low is four-sourced on the figure and zero-sourced on the time, so the claim that it printed before 01:00 UTC rests on it appearing in a 10:00 JST summary and not in a 09:04 JST one — an inference from two publication stamps, not a tick record. The 156.06 support comes from a single publisher’s daily technical note and always did; if that reading is wrong, the 2.0-pip hold is wrong with it. The Tokyo-fixing attribution is two publishers’ characterisation of a flow neither of them measured, and we have no way to weigh it against the level. The 157.16 London high and the 156.64 spot are each single-sourced. The 21-day at 157.88 is two-sourced; the 200-day values are one reader each and, as above, we no longer think the spread between them means what we have been publishing. And the thing we are most confident about is a defect in our own process rather than anything about the yen.
Related
- FX events calendar — the American block at 13:45 and 14:00 UTC, and the New York option cut.
- Signals
- EA presets
