Global Rates
Bank of Japan September
98.35c (was 88.5c)
A TEN-POINT move in one session, the largest single-day repricing on this board since we began tracking it. The JGB 10-year touching 3.00% for the first time since 1996 did the work in cash before the meeting.
10Y Treasury
4.79% (Sep 1 SETTLE)
Treasury par yield curve, primary source. Fifth consecutive higher close; 2-year 4.39%, 30-year 5.27%. ONE basis point from the 4.80% touch. Vendors reported 4.77-4.81 intraday and the contract is unresolved, so we do not claim a print.
pos-017, one day after we sold
94.25c vs our 71.69c fill
Sold Tuesday on a published gate into a book with ten shares on top. Today's 91c bid carries 271. Foregone $29.27 now, $42.91 at resolution. Correct process, bad outcome, printed above what worked.
Brent crude
$94.65 SETTLE (+4.6%)
WTI $90.22, +5.2%. US strikes on Larak Island and Trump's explicit threat to Kharg Island, through which roughly 90% of Iran's crude exports flow. Yesterday's unverifiable-settle flag comes OFF - this is a settle with a change attached.
Central bank board
BoJ 98.35 / ECB 98.95 / Fed 59.5
All three September meetings priced to tighten, all LIVE. The Fed leg against a 65-68% FedWatch is an 8.5-point gap - still under our 10-point bar and a fourth correlated expression. Refused for a third session.
Retired barrier model
7.8pp gap, was 11.1pp
The 5.00% diagnostic narrowed sharply. Yesterday's reinstatement rule requires within 5 points for three consecutive sessions; session one FAILS at 7.8. The model does not get its job back for one good day.
The central-bank board has stopped being a board and started being a single trade. The Bank of Japan's September meeting now prices 98.35c for a 25 basis point increase, up from 88.5c yesterday - a ten-point move in one session, and the largest single-day repricing on this board since we started tracking it. The ECB's September meeting prices 98.95c. The Fed's prices 59.5c for a hike, with CME FedWatch reported between 65 and 68 percent. Three of the world's four largest central banks are now priced as near-certainties or coin-flips to TIGHTEN at meetings inside the next fortnight, and the cash markets got there first: the JGB 10-year touched 3.00% on Tuesday for the first time since 1996, the Bund reached 3.3% for the first time since May 2011, and the US 10-year settled 4.79% - its fifth consecutive higher close and one basis point from the level our pos-017 was written on. That last number is the reason this letter leads with a position we sold. We closed pos-017 on Tuesday morning at a volume-weighted 71.69c because a published take-profit fired at an 82c bid with ten shares behind it. It trades 94.25c now, with a 91c bid 271 shares deep. The gate has cost $29.27 at today's price and $42.91 if the touch resolves. We publish that above the two positions that are working. The proximate driver stayed where it was on Tuesday and got worse: US forces struck Iranian installations on Larak Island, President Trump extended explicit threats to Kharg Island - the terminal through which roughly 90% of Iran's crude exports flow - and Brent settled $94.65, up 4.6%, with WTI at $90.22, up 5.2%. An energy shock of that shape does not give any of these three committees a disinflation argument, which is why the board looks the way it does this morning.
Today's Market Moves
Bank of Japan September 25bp increase
88.5%→98.35%+9.85pp
Bid 98.3 / ask 98.4 on $149K; no-change collapsed to 1.05c from 11.5c. The largest one-session move this board has shown. The JGB 10-year at 3.00% - a level untouched since 1996 - is the cash market having already paid for the hike. No edge left at 98.35 and we would not add a fifth tightening bet for one anyway.
10Y Touches 4.8% Before 2027 (pos-017, CLOSED Tue)
87.5%→94.25%+6.75pp
Not ours. Bid 91 / ask 97.4, hourly high 97.25c. The 91c bid carries 271 shares against ten at 82c on Tuesday, which is the whole difference between the fill we got and the one available now. $29.27 foregone at today's bid, $42.91 at resolution.
10Y Touches 5.00% Before 2027 (diagnostic)
33.5%→43.0%+9.5pp
Bid 37 / ask 49. The retired model says 50.8 on a 21.0bp barrier over 82 sessions with a 36.22bp horizon sigma. Gap 7.8pp, narrowed from 11.1. Session one of the three-session, five-point reinstatement test FAILS. Diagnostic only, and it stays that way.
Fed September 25bp increase
56.5%→59.5%+3.0pp
Bid 59 / ask 60 on $15.8M; no-change 40.5c. FedWatch reported 65-68% on Sep 1 - we use 68 and flag the range rather than picking the number that makes the gap look best. 8.5 points, under our bar for a third session.
ECB September 25bp increase
98.55%→98.95%+0.4pp
Bid 98.8 / ask 99.1. Effectively resolved. The Bund at 3.3%, French yields at 2008 highs and peripheral spreads rising with the core is the same trade already expressed in cash, and far more cheaply.
Fed Rate Hike in 2026 (pos-010)
71.5%→71.5%0.0pp
Flat on the mid but the hourly printed 72.5c at 06:00 UTC, so the 72c trim fired a SECOND time in 26 hours. Honoured at the 71c bid (9,092 shares of depth) for +$7.27, two points below our own fair value of 73. Hysteresis rule added: re-arms only below 68c.
Screening Table
| # | Market | Expiry | Market Price | Fair Value | Gap (pp) | Direction | Volume | Confidence |
|---|---|---|---|---|---|---|---|---|
| 1 | Bank of Japan September 25bp increase | Sep 18 | 98.35% | 98% | -0.35pp | NO POSITION - a ten-point one-session repricing that the JGB at 3.00% had already paid for; no edge and no appetite for a fifth tightening bet | $$149K on the leg | 4/10 |
| 2 | Fed September 25bp increase | Sep 16 | 59.5% | 68% | +8.5pp | NO POSITION - 8.5pp to a 65-68% FedWatch, under the 10pp bar and a fourth correlated expression; refused for a third session | $$15.8M on the leg | 3/10 |
| 3 | 10Y Touches 5.00% Before 2027 | Dec 31 | 43.0% | retired% | 0pp | NO POSITION - the diagnostic that retired our model; gap narrowed 11.1 -> 7.8pp and session one of the reinstatement test still fails | $$85.4K on the leg | 5/10 |
| 4 | ECB September 25bp increase | Sep 10 | 98.95% | 98% | -0.95pp | NO POSITION - resolved in all but name; the Bund at a fifteen-year high is the same trade in cash | $$121K on the leg | 3/10 |
| 5 | 10Y Touches 4.8% Before 2027 (pos-017) | Dec 31 | 94.25% | n/a% | 0pp | CLOSED Tuesday at 71.69c - trades 94.25c today; $29.27 foregone at the bid, $42.91 at resolution, published above what worked | $$62.7K on the leg | 5/10 |
Top 5 Opportunities
1
Bank of Japan September - a ten-point repricing that the 1996 print had already paid for — NO POSITION
↑ BUY YES-0.35pp
Market price
98.35%
Fair value
98%
Gap: -0.35pp
The September BoJ meeting went 88.5c to 98.35c in a single session for a 25 basis point increase, with the no-change leg collapsing from 11.5c to 1.05c. That is the largest one-day move this board has produced since we began tracking it, and it happened without a meeting, a statement or a leak. What happened instead was in the cash market: the JGB 10-year touched 3.00% on Tuesday, a level it has not seen since 1996, and the prediction market simply caught up to a price that Japanese government bonds had already set. This is worth dwelling on because it is the second time in three sessions that the cash market has led the prediction market by hours on something we track. It is also why we have no position. Our fair value is 98 and the ask is 98.4; there is nothing to collect. And the broader point about Japan matters more than this contract does. Japanese institutions have been the marginal buyer of long-dated foreign paper for two decades. A domestic 10-year at 3% removes the reason to own hedged Bunds or Treasuries at the spreads currently on offer, and that repatriation channel is a substantial part of why the Bund printed 3.3% and the US 10-year 4.79% in the same week.
▵ Bull case
- The JGB 10-year at 3.00% for the first time since 1996. The cash market has priced the hike more emphatically than the prediction market can.
- Brent settled $94.65, up 4.6%, with Kharg Island under explicit threat. Japan imports essentially all of its crude.
- The ECB is at 98.95c and the Fed at 59.5c for the same month. A BoJ that declined to move would be the outlier on a board that has stopped disagreeing.
- 98.35c pays roughly 1.7% over sixteen days if it resolves as priced.
▿ Bear case
- Our fair value is 98 against a 98.4 ask. There is no edge before any portfolio consideration is applied.
- It would be a fifth simultaneous expression of global tightening in a four-position book. Correlation is the risk we have named as our largest for a month and today it would be worse, not better.
- The BoJ has declined to validate confident market pricing before, and 1.05c is a thin price for that history.
- $149K of volume on the leg. Tuesday and today both taught us what happens to a real fill in a thin book, twice, on two different positions.
2
The 4.8% touch is one basis point away and we sold it on Tuesday — YES
↑ BUY YES0pp
Market price
94.25%
Fair value
0%
Gap: 0pp
The US 10-year settled 4.79% on Tuesday, its fifth consecutive higher close, which puts it one basis point from the level pos-017 resolves on. We sold that position on Tuesday morning at a volume-weighted 71.69c because a take-profit written on a 66c bid fired at an 82c bid. The contract trades 94.25c this morning and printed 97.25c overnight; the 91c bid now carries 271 shares against the ten that sat behind Tuesday's 82c. Our 151.5152 shares would realise $137.88 today rather than the $108.61 we took. The gate has cost $29.27 and will have cost $42.91 if the touch prints. We are printing this above the positions that are working, because a letter that reports its gates only when they pay has no process to describe. We are also not changing the decision retroactively. The rule was set when this position's fair value was 60, published every day for a fortnight, and honoured on the morning it fired into the thinnest book we had seen on it. What we will say is that a fixed take-profit which never tracks fair value will systematically sell winners early, and we left ours at 66c while our own arithmetic moved to 76.8 and then higher. That is a named flaw now rather than an unexamined default.
▵ Bull case
- The exit was the rule working. A take-profit honoured only when subsequent prices validate it is not a rule.
- Tuesday's book had ten shares at 82c. The 271-share bid at 91c was not a choice available to us that morning.
- +$83.61 on $25 remains the second-largest winner this book has produced, and 334% does not become a failure because a larger number existed.
- Closing removed the largest correlated hawkish-rates bet in the book, which is diversification we could not otherwise achieve.
▿ Bear case
- $29.27 at today's bid, $42.91 at resolution, on a $25 stake. No framing shrinks those.
- The 66c trigger was set at a fair value of 60 and never revisited as the thesis strengthened. We knew fixed gates sell winners early and left it anyway.
- Our published fair value was frozen at 61 by a rule about a different market while the arithmetic said 76.8. Without the freeze the gate would have looked obviously stale against our own number.
- The barrier model we retired on Tuesday for five consistency failures was closer to right on this leg than the market was. Retiring it the same morning we sold is a coincidence worth sitting with.
3
Fed September - 8.5 points to FedWatch, refused for a third session — NO POSITION
↑ BUY YES+8.5pp
Market price
59.5%
Fair value
68%
Gap: +8.5pp
CME FedWatch was reported between 65 and 68 percent for a 25 basis point September increase on Sep 1. We use 68 and flag the range, rather than quietly selecting whichever number makes our gap look best - the difference matters, because at 65 the gap is 5.5 points and at 68 it is 8.5, and neither clears our bar. Polymarket trades 59.5c with a one-cent spread on $15.8M. This is the third consecutive session we have declined it, and the reasons have not changed: it sits below our published 10-point entry bar, and it would be a fourth expression of hawkish-rates risk in a four-position book. What HAS changed is that the correlation argument is weaker than it was. We closed pos-017 on Tuesday and trimmed pos-010 twice, taking roughly $100 of stake out of the rates complex in two sessions. A reader is entitled to notice that the constraint we keep citing is loosening while the trade we keep refusing stays attractive. Our answer is that a bar which bends the week its binding constraint relaxes was never a bar, and that we would rather miss this than explain that.
▵ Bull case
- 8.5 points on a $15.8M contract with a one-cent spread, and prediction markets have historically converged toward futures-implied odds into FOMC dates.
- Cleveland's first September core PCE nowcast is 3.49%, accelerating from 3.40%. Inflation is getting worse three weeks before the meeting.
- ISM manufacturing 54.6%, GDPNow 4.8%, Brent settled $94.65. There is no growth or energy argument against tightening.
- The book's rates correlation is genuinely lower than it was on Friday - two of the constraints we cited when first refusing this have partly resolved.
▿ Bear case
- 8.5 is below 10. Moving a published bar because the trade looks good is how a process becomes a narrative.
- The FedWatch reading spans 65-68 across sources and we could not parse CME's own tool this morning. We are comparing a market price to a number we did not verify at source.
- It would still be a fourth correlated expression, and correlation risk does not disappear because it merely got smaller.
- The August employment report is Friday. Consensus +60,000 with unemployment ticking to 4.2%, and July payrolls FELL 23,000. A weak print reprices this leg violently before the meeting.
4
10Y Touches 5.00% - the retired model is improving and does not get its job back — NO POSITION
↑ BUY YES0pp
Market price
43.0%
Fair value
0%
Gap: 0pp
On Tuesday we retired the barrier model to diagnostic-only after it failed its consistency check on this leg five consecutive times, and we wrote a reinstatement rule at the same time: the check must reproduce within five points for three consecutive sessions. Today the model returns 50.8 on a 21.0 basis point barrier over 82 sessions against a 43.0c market. The gap is 7.8 points, narrowed sharply from 11.1. That is the model behaving better than it has in a fortnight, and it is session one of three, and 7.8 is not within five. So it fails, and the model stays retired. We are stating that plainly because the temptation runs the other way: the model was closer to right than the market on the neighbouring 4.8% leg, the gap here is closing fast, and we have an obvious incentive to reinstate a tool that would have told us to hold pos-017. Yesterday's rule was written before we knew any of that. It binds.
▵ Bull case
- The gap narrowed from 11.1 to 7.8 in one session, the first sustained improvement since the check started failing.
- The model was closer to right than the market on the 4.8% leg, which is now one basis point from resolving.
- The 10-year settled 4.79% with the global long end at multi-decade highs. The path to 5.00% is more credible than at any point this year.
- The Treasury buyback window that produced our 0.9045 haircut expires Nov 4 while the barrier runs to Dec 31 - two months with less official support.
▿ Bear case
- 7.8 is not within 5, and session one of three is not three sessions. The rule was written yesterday, before we could know today's numbers.
- We have a transparent incentive to reinstate this model right now, which is exactly the condition under which a pre-committed rule earns its keep.
- The spread is 12 points wide - bid 37, ask 49 - and this week has twice shown what a wide book does to a real fill.
- It would be a further correlated rates bet in a book we spent two sessions de-correlating.
5
ECB September - resolved in all but name, and the Bund is the cheaper version — NO POSITION
↑ BUY YES-0.95pp
Market price
98.95%
Fair value
98%
Gap: -0.95pp
98.95c eight days from the meeting is a market that has stopped asking a question. The cash market agrees and got there first: the Bund 10-year at 3.3% is the highest since May 2011, French yields the highest since November 2008, Dutch at fifteen-year highs, with Italian and Spanish spreads rising alongside the core rather than against it. That last detail is the one worth carrying forward. A periphery selling off in sympathy with the core is not how a market prices a central bank fighting demand-side inflation; it is how a market prices term premium and fiscal supply. Six weeks ago this leg was an inflation trade and it has quietly become a duration trade, which changes how we would read a European surprise even though we will never hold this contract. There is 1.05 cents of upside here over eight days, against a 98.95-cent downside, in a market where $121K of volume means our own ticket moves the price.
▵ Bull case
- ECB minutes point to a further increase as likely necessary, and the Bund at a fifteen-year high is the curve agreeing in cash.
- Brent settled $94.65 with Kharg Island threatened - a euro-area energy shock with no domestic offset.
- The BoJ is at 98.35c and the Fed at 59.5c for the same month. A board this aligned rarely breaks.
- 1.05 cents over eight days annualises to a number a money-market desk would look at twice.
▿ Bear case
- 1.05 cents of upside against a 98.95-cent downside. We would refuse that shape anywhere else and we refuse it here.
- $121K of volume means our own size moves the price against us, which this week has demonstrated twice on two different positions.
- A fifth simultaneous tightening bet in a four-position book.
- Pricing a 98.95c contract requires being right about the 1.05c, not the 98.95c, and we have no European informational edge whatsoever.