Yesterday morning this desk wrote that no option expiry ladder for 14 September existed at 06:20 UTC, and concluded that London would therefore open on Friday’s map. The first half was true. The second half was wrong. InvestingLive published the 14 September ladder at 06:58:44 UTC — thirty-eight minutes after we looked, and two minutes before the London open. London did not trade Friday’s map. It got Monday’s, with a hundred and twenty seconds to read it. This morning at 06:30 UTC there is again no ladder for today, and we now know that tells us nothing about whether there will be one at 06:58.
What we got wrong, and it was the inference, not the fact
The observation was correct: at 06:20 UTC on 14 September the publisher’s order index carried nothing for that day. The error was treating an absence at 06:20 as a fact about 07:00. We published a consequence — that the London open would be traded against a three-day-old strike map — and that consequence did not happen, because the page appeared in the thirty-eight-minute gap between our fetch and the bell.
This is a small mistake with a specific shape, and it is worth naming because this channel makes it often enough to be a category. A negative observation about a data source is a statement about one moment. It becomes a statement about a session only if you also know the source’s publication cadence. We did not know it. We asserted the consequence anyway. That is the whole error.
So this morning, instead of repeating the assertion, we went and measured the cadence.
Four publication times, spread across more than eight hours
From the publisher’s own index and article timestamps, here is when the last four New York cut ladders actually appeared:
- The first 11 September ladder: 10 September, 22:19 GMT — the evening before, comfortably ahead of any European session.
- The second 11 September ladder: 11 September, 05:01 GMT — a separate article, different byline, materially different notionals. This is the pair this desk audited yesterday.
- The 14 September ladder: 14 September, 06:58:44 UTC — two minutes before the London open.
- The 15 September ladder: not published at 06:19 UTC, and still not published at 06:30 UTC, checked twice eleven minutes apart.
From 22:19 the previous evening to after 06:30 the following morning is a spread of more than eight hours. That is not a cadence. There is no hour at which you can schedule a fetch and expect the page to be there, and no hour at which its absence means anything.
One correction to our own arithmetic while we are here. Yesterday we described the second 11 September ladder as arriving “seven hours later” than the first. The index timestamps are 22:19 and 05:01, which is six hours and forty-two minutes. The argument does not change and the number should have been right.
And when the page does exist, the numbers may not
There is a second failure mode stacked on top of the timing one, and it is worse. The 14 September ladder — the one that did arrive, two minutes before the bell — does not yield its notionals to a fetch. The strikes and amounts live inside an image file embedded in the page. What extracts as text is a prose mention that EUR/USD has a notable expiry at 1.1550 with larger ones at 1.1500 and 1.1600, and no dollar amounts anywhere.
So the realistic worst case for an automated pre-London process is this: the page does not exist when you check, it appears while you are not looking, and when you do read it the quantities that make the strikes worth caring about are in a picture. Three independent things have to go right, at least one of which is outside your control, on a source that has already been shown to run two contradictory versions of the same day’s ladder under two bylines.
This is the third article this desk has written about this one dataset in five sessions. The first published the ladder as fact. The second showed the publisher carried two versions of it. This one shows you cannot reliably get either version in time. We are not telling you to stop reading it. We are telling you that the confidence we attached to the $2.86 billion figure on Friday was not warranted by the pipeline that delivered it, and that is our error rather than the publisher’s.
What a pre-London process should actually do about it
Three changes, none of them clever.
First, stop treating the ladder as a scheduled input. Anything that arrives somewhere in an eight-hour window cannot be a gate in a process that has to make a decision at a fixed time. If your London-open logic waits for it, your London-open logic waits indefinitely on some days. Poll it, use it when it is there, and have the session behave identically when it is not.
Second, record the timestamp on the ladder you are using, and let position size see that number. A ladder published at 22:19 the previous evening and a ladder published at 06:58 today are not the same quality of information, and neither is a three-day-old one. The staleness belongs in the sizing rule, exactly as it does with the Tuesday-close measurement date on positioning data, which this desk has been saying for a week and is now saying about its own favourite source.
Third, size on the strike, not on the notional. The strike survives a bad fetch. It is in the prose, it is in the text, it does not live in a picture, and it is the part you would actually act on — a pin risk around a round number does not need the billions attached to it to change how you would place a stop into the New York cut. The notional is a confidence weight, not a signal, and it is the part most likely to be missing.
Where spot is while you wait
USD/JPY spent the Tokyo session in a 68-pip box, 154.21 to 154.89, and was at 154.80 at 15:12 JST, pressing the top of it into the London handover. That follows an overnight range of 155.00 to 153.99 — a hundred and one pips out and back, which this desk covered this morning and which had nothing Japanese in it. Two publishers give the same session extremes today, which is a clean read.
The two vendor maps for today disagree in the way they have all week, and the disagreement is the useful part. Gaitame gives a forecast range of 153.500 to 154.900, resistance at 155.00 and support at 153.00. Fisco gives the same 155.00 resistance but puts support at 152.00, with 150.00 in view below it. Both name 155.00; they are a full figure apart underneath. If you are sizing a downside stop this morning, you are choosing between two published floors that are a hundred pips apart, and neither publisher is more entitled to your trust than the other.
EUR/USD ran 1.1533 to 1.1552 in Tokyo. EUR/JPY 178.14 to 178.68. GBP/USD came into the London open grinding lower near 1.3490 after a five-week low at 1.3464, with the UK labour market print at 06:00 and the Bank of England on Thursday at 11:00 UTC. And the 14 September ladder’s one extractable strike, EUR/USD 1.1550, sits inside this morning’s Tokyo range — which is either mildly interesting or a coincidence, and on a ladder that is a day out of date we are calling it a coincidence.
What this does not tell you
Four publication timestamps are four observations, not a distribution. We can say the spread is wide because we have seen it be wide; we cannot tell you the probability that today’s ladder lands before 07:00, and we are not going to dress four data points up as a model. If this matters to your process, log the arrival time every day for a month and you will have something we do not.
We also do not know why the timing varies. A different author on different days is the obvious guess and it is only a guess — the 11 September pair did carry different bylines. Whether that explains the 22:19 versus the 06:58 we have not established, and we are not going to name a person as the cause of a publication schedule on the strength of two bylines.
And we cannot tell you what is actually on today’s board, because at the time of writing it does not exist to be read. Everything above is about the pipeline, not about the strikes. If the ladder appears at 06:58 again, the levels in it will be news and this article will not have contained them.
