The Number That Would Settle It Is Three Numbers, Three Pips Apart

We asked for one number and we were sent three

Three readers give the dollar–yen two-hundred-day moving average as follows. A Tokyo broker’s morning note, filed 23:24 UTC last night, puts it at 158.53 and makes it the day’s single focus. A Japanese technical column filed at 07:40 UTC today puts it at 158.52. MUFG, relayed at 10:07 UTC, puts it at 158.50 and says in as many words that today’s 158.51 high “is just at the 200-day moving average level”.

So the band is 158.50 to 158.53. Three pips wide. Tokyo’s high of 158.51 sits inside it and is equal to none of the three: one pip above the lowest reading, two pips below the highest.

This is not a vendor error and we are not going to present it as one. A moving average has no issuing institution. It depends on which closing convention you use, which data vendor’s closes you use, and whether the series is a simple or an exponential average over two hundred sessions that include a Tokyo holiday or two. Three publishers computing it honestly from three slightly different close series will land three pips apart, and all three are right about their own series. There is nothing to settle. That is the finding.

Our own published level is one of the three, which makes scoring it impossible

Here is the part that costs us something. The resistance level this desk published three days ago was 158.50. MUFG’s two-hundred-day moving average is 158.50. They are the same number to the pip.

Yesterday we declined to claim the one-pip overshoot as a vindication, on the grounds that we could not distinguish our level from the average. That refusal now looks better than we knew. “Our level held” and “the two-hundred-day held” are not two competing claims about today’s session. On MUFG’s reading they are the same sentence, written twice, and we published one of them without knowing it duplicated the other.

It gets denser. The same publisher relaying MUFG’s 158.50 filed its own technical note at 05:30 UTC placing the fifty per cent retracement of the move at 158.52 — the identical value to the technical column’s two-hundred-day. So inside a three-pip band you have one resting-order level that we put there, three computations of the same average, and a Fibonacci retracement. Five candidate causes, three pips, one touch.

Why this matters more for the stop than for the story

Yesterday’s piece made a decay-rate argument: an order level sits until it is filled or pulled, a moving average moves every day, and an expiry ceases to exist at the New York cut, so a stop placed above a level on an order-flow theory drifts out of position if the level was actually an average. That argument survives. What has changed is that it now has a measurement problem sitting on top of it.

You cannot size the drift if you cannot identify the cause to better than three pips, and three pips is three times the precision of the one-pip fit we were tempted to claim. A short stop placed at 158.55 is above all three readings today. By Friday, if the pair is still firm, the averages will have moved up and it may be above none of them, and nothing in your terminal will tell you that the reason you chose the level has walked away from it. The honest instruction is not “trust the average” or “trust the order level”. It is: in this three-pip band, place the stop on your own loss tolerance, because none of the five reasons available to you is measurable enough to carry it.

One more piece of drift, recorded because it is cheap to check and nobody does: the same publisher gave the hundred-day average as 159.55 at 02:03 UTC and 159.53 at 05:30 UTC today. Two pips in three and a half hours, from one source, on one indicator, on one day. That is the scale of precision an intraday technical level actually has.

London did what Tokyo did not, and it did it downward

Yesterday we left a question open: whether London and New York would produce more than Tokyo’s forty-one pips, and whether 158.51 would hold as the day’s high. Both halves now have answers, and the second one is the less interesting.

158.51 held. MUFG at 10:07 UTC still described it as “the high today”, and nothing we read between 06:00 and 11:30 UTC puts a higher print on the board. The resistance side of our ladder therefore scores: 158.00 and 158.40 gave way earlier in the week, 158.50 was exceeded by one pip and rejected, and 159.00 — forty-nine pips above the high — has still never been approached.

The support side is where the session went. London took the pair to 157.85 at about 08:57 UTC, twenty-five pips below Tokyo’s low, before it recovered to 158.12 and then 158.17 by 09:20 UTC. That makes the day’s range 66 pips against Tokyo’s 41 — 1.61 times as much. Our own 158.00 broke downward by fifteen pips. The next rung at 157.50 was approached to within thirty-five pips and left untested, so it is still an unscored number and we are not going to pretend otherwise.

Note what this does to a benchmark we published two days ago. We called twenty-eight pips a quiet day with an event in it. Tokyo gave forty-one with no event. London gave sixty-six by late morning, also with no scheduled Japanese or European print of consequence after 06:00 UTC. The quiet-day number is retired for good; it was measuring a quiet hour, not a quiet day.

The move that broke 158.00 was not a Japan story

This is an Asia desk, so we will say the uncomfortable thing about our own session. The largest single move in the dollar–yen today was not made in Tokyo and had nothing to do with Japan.

Two Japanese-language London reports attribute the dip under 158.00 to euro selling spilling across the board. The euro was the day’s driver: euro–dollar fell to 1.1198 by 09:20 UTC, fifty-six to sixty-five pips off the Tokyo high depending on which print you take, and euro–yen to 177.13, some 117 pips off its Tokyo high. The cause, as our Macro colleagues set out separately today, is French government paper being sold again on fiscal and political uncertainty.

Operationally that is a correlated-exposure problem, not a yen problem. If you are short the yen against the dollar and you treat that as a Japan trade sized on Bank of Japan pricing, you were handed twenty-five pips of adverse movement this morning generated by a French bond market. The October rate-rise probability on the Japanese side has not been the operative variable today at any point. For completeness, the probability doing the work on the dollar side is the Fed’s: CME FedWatch was at about twenty-two per cent for an October increase as of 08:43 UTC, against the twenty-four per cent implied by the seventy-six per cent hold reading this desk has been carrying from a different feed. A two-point gap between two feeds, which is the narrowest agreement we have recorded on a central-bank probability in a month.

What this does not tell you

Whether 158.51 survives New York is unknown at the time of writing. We are publishing at 11:52 UTC, before the United States has opened, before a ten-year note reopening at 17:00 UTC and before the September Federal Open Market Committee minutes at 18:00 UTC. If the pair takes out 158.53 this evening, the whole three-pip band becomes a footnote and the next real question is 159.00.

We did not reach an issuing institution for any of the three moving-average values, because there is no such institution. All three are secondary by construction and we have labelled each with its publisher and its filing time rather than picking one. We have not verified that all three are simple rather than exponential averages, and that alone could account for the spread.

We also cut a figure this morning. A Japanese-language note carried an “October rate increase” probability of roughly twenty per cent in a paragraph otherwise about the Federal Open Market Committee minutes, and we could not establish from the text whether it referred to the Fed or the Bank of Japan. Two central banks, one ambiguous antecedent, so the number is not in this article.

And none of the above is a view on direction. We have told you that five reasons are stacked in three pips and that none of them is measurable to better than the band. That is an argument about how much to risk across the 17:00 and 18:00 UTC windows, not about which way to lean into them.

Related

• FX events calendar — the 17:00 and 18:00 UTC windows for today
• Signals
• EA presets