66.8, Just Under 80, or 88. Three Feeds, One Bank of Japan Meeting.

Ueda said the hard part out loud, and it was not about October

Two readers carry the parliamentary remarks. The first, timed 02:18 UTC, has him saying underlying inflation is “quite close” to target, warning about “failing to adjust our policy rate appropriately”, and attributing capital expenditure and export strength to “very strong global AI demand”. The second, timed 06:52 UTC, reports via Danske Bank that he warned that failing to raise rates at an appropriate pace could push long-term bond yields higher. The two readings agree on substance and neither is a headline scrape; both are bodies.

Read as a reaction function rather than a signal, this is a governor describing the cost of being late rather than the case for being early. That is not an October argument. The policy rate is 1.25 per cent after September’s move, and the expectation across this morning’s sweep is a hold. What Ueda added is the asymmetry: in his framing the risk of waiting is priced in the bond market, not in the inflation print. If you trade the yen around policy, that sentence is worth more than any probability below, because it tells you which way the bank would rather be wrong.

Three feeds, three December numbers, and 21 points between the ends

Here is the board as of this morning, with what each figure actually measures:

• Just under 80 per cent for a December hike — a Tokyo house’s weekly outlook, published 03:16 UTC today. Stated qualitatively, in words, with no feed named.
• 88 per cent for action by the December meeting — attributed to overnight index swap pricing, published 13 October. Six days old, and the word is “by”, not “at”.
• 66.8 per cent for a hike at the December meeting, with 58.1 per cent cumulative probability that the rate is higher than today — a dedicated central-bank probability page, data stamped 16 October.

The ends are 21.2 points apart. If you insist on comparing like with like and set the 88 per cent for action by December against the 58.1 per cent cumulative, the gap widens to 29.9 points. That is the whole range between nearly certain and a coin flip with a lean, for a binary event eleven days from its first opportunity.

This desk has now gone seven slots without a single published Fed probability in the sweep and has three competing Bank of Japan figures in one morning. We are not going to pick one. The spread is your error bar, and any inference that needs the precise level is weaker than it reads.

One of the three prices a cut, sizes its own hike from its own probability, and dates the meeting wrong

The 66.8 per cent page deserves a paragraph on its own, because its internals are checkable and they do not hold up. For the October meeting it shows no chance of a hike and a 13.1 per cent chance of a cut — against every other reader in the sweep, which has October as a hold with a small hike tail. It also carries a policy rate of 1.25 per cent in one section and 0.50 and 1.04 per cent in others, unreconciled.

Then the arithmetic. It reports the December move as plus 16.7 basis points. Sixteen point seven is 66.8 per cent of 25, which is its own stated probability times a full hike. The expected move is not an independent measurement; it is the probability restated. That is fine as presentation and misleading if you read the two fields as corroborating each other.

Finally the date: it places the October decision on the 28th. Two separate readers this morning put the meeting’s conclusion on 30 October. We could not reach the bank’s own 2026 schedule to settle it, and this is where we nearly manufactured a finding: the schedule page that search returned is the 2027 calendar, whose October meeting runs 28 and 29. We offer that as a plausible mechanism for the error and not as a fact, because we did not see the aggregator’s source. The lesson is the one this desk keeps relearning, and it cut in our own direction this time: check the year on anything a schedule page hands you, including a central bank’s own.

The consequence is a sizing rule, and it is not a view on the yen

Dollar-yen spent today inside about 40 pips, from the 157.70 area in the Tokyo morning to around 158.10 after London dipped and bought it back. Forty pips, on a morning when the governor told parliament that being late is the expensive error. That non-reaction is the most informative thing on the screen: the market is not trading the October meeting, and it is not confident enough about December to move either.

So the rule for the next eleven days is about exposure windows, not direction. Japan’s national CPI lands Friday at 23:30 UTC, with one reader’s core forecast at plus 2.1 per cent year on year against plus 1.7 prior — a 40 basis point acceleration that, if it prints, moves the December question before any central banker speaks again. The decision itself is the end of the month. Between now and Friday there is no scheduled Japanese event large enough to resolve a 21-point probability spread, which argues for being in the market at normal size and out of it across the Friday window, rather than the reverse.

We will not tell you whether the bank hikes. We will tell you that three feeds cannot agree on how likely it is, that one of them cannot agree with itself about the current policy rate, and that the governor has told you which mistake he is more afraid of.

What this does not tell you

It does not tell you what the Bank of Japan will do in October or December. It does not give you a fourth probability reader, which is what a 21-point spread deserves and what we did not find in the time available. It does not establish the bank’s 2026 meeting dates at the issuer — two secondary readers give 30 October and we could not confirm it at source, so treat the date as well-attested rather than primary. It does not verify the 88 per cent figure against the swap curve ourselves; it is one publisher’s reading of overnight index swaps and it is six days old, which on a probability is a long time. It does not reproduce Ueda’s remarks in full, only the fragments both readers carry. And the 40-pip range is today’s session on two Tokyo readers, not an official range.

Related

• FX events calendar — the Friday 23:30 UTC Japanese CPI window and the month-end decision.
• Signals — how a policy window is sized when the probability has no agreed level.
• EA presets — news-window configuration for central-bank days.
• When the Bank of Japan was already priced — the same question at a previous decision.
• The gap closed before the instrument meant to price it opened — this morning’s companion note on the same session.