Global Rates
pos-017 fair value
61 (FROZEN)
Model says 65 on a 13.6bp barrier after the 10Y settled 4.664%. The Aug 25 standing rule blocks the increase - the 5.00% check gives 34.0 vs a 17.0c market, WIDER than yesterday
US 10Y
4.664%
+2.5bp, Aug 26 SETTLE, verified. 30Y 5.186% (+1.2bp). Armed trim rule needs two consecutive settles below 4.60% - Aug 25 was 4.639%, Aug 26 was 4.664%, NOT triggered
Bund / JGB / gilt
3.2550% / 2.8950% / 5.0676%
CARRIED from Aug 25 for a SECOND session - yield feed failed at source again. Now two sessions stale and we will not infer them from anything
ECB / BoJ / Fed Sep
96.75c / 86.0c / 32.5c
ALL CARRIED from Aug 26 - Polymarket unreachable. These are last-known prices, not current ones
July PCE (released)
3.7% YoY / core 3.3%
Above the 3.6% consensus we published. Cleveland nowcast had 3.65% and core 3.29% - closer than consensus on every line, and slightly low on headline
USD/JPY
159.30
LIVE Aug 27 quote. Aug 25 settle was 159.22. Still just under the level that triggered joint intervention this month, and drifting the wrong way for the BoJ
Two things happened to this letter today and both of them are about method rather than markets. The first is that our own standing rule blocked a favourable revision. The US 10-year settled at 4.664%, up 2.5 basis points, narrowing the pos-017 barrier to 13.6 basis points and lifting the model's answer to 65 from yesterday's 61. We are not taking the upgrade, because the rule we wrote on Aug 25 says no fair-value increase until the 5.00% consistency check reproduces, and today that check got worse rather than better: 34.0 from the model against a 17.0c market, a 17pp gap against yesterday's 14.1pp. The second is a data outage. Polymarket has been unreachable for this entire session - 403 responses from the API, blank pages on the site. Every prediction-market price in this letter is carried from Aug 26 and labelled. That means the central-bank board below is a day old. The ECB September leg, the BoJ September leg and the Fed September leg are all last-known prices, not current ones, and the arithmetic we ran against them yesterday cannot be re-run today against anything fresh. We are publishing with the gap named rather than skipping, because the non-market content is real: July PCE printed at 3.7% headline and 3.3% core, above the consensus we published and almost exactly on the Cleveland Fed nowcast this letter uses as its anchor. GDPNow went from 4.0% to 4.6%, verified at source. Durable goods rose 1.1% against a 0.5% forecast. Inflation-adjusted consumer spending was flat. And the sovereign yield feed that failed yesterday failed again today, so Bund, JGB and gilt levels remain carried from Aug 25 and are now two sessions stale. We flag that rather than quietly reprinting them as though they were fresh.
Today's Market Moves
10Y Touches 4.8% Before 2027 (pos-017)
65.5%→65.5%0.0pp
CARRIED from Aug 26 - Polymarket unreachable all session. Price stale, arithmetic fresh. 10Y settle 4.664% gives a 13.6bp barrier over 90 sessions with horizon sigma 37.95bp: 72.0% raw, 65.1 after haircut. FV would be 65. It stays 61 because the Aug 25 rule blocks increases while the 5.00% check fails, and with the comparison leg also carried we cannot even run that check cleanly. An unresolvable check is not a passed check.
10Y Touches 5.00% Before 2027 (watchlist)
17.0%→17.0%0.0pp
CARRIED from Aug 26 - Polymarket unreachable all session. Model 37.6% raw, 34.0 after haircut, against a carried 17.0c. Nominal gap 17pp, wider than yesterday's 14.1pp and the third consecutive session this leg is refused. The widening is entirely because the 10-year rose and the barrier narrowed - the market leg has not been re-observed at all. That is precisely why we will not treat a bigger number as a better opportunity.
ECB Interest Rates September 2026 - 25bp increase
96.75%→96.75%0.0pp
CARRIED from Aug 26 - Polymarket unreachable all session. Yesterday we argued this leg had stopped pricing economics and started pricing execution, after it ROSE on a 10% fall in imported energy. Today's US inflation print is mildly supportive of that reading - hot core with no obvious euro-area read-across - but we cannot test it, because we have no fresh price. The reading stays a reading. Decision Sep 10, euro-area flash HICP Aug 31.
Bank of Japan September - 25bp increase
86.0%→86.0%0.0pp
CARRIED from Aug 26 - Polymarket unreachable all session. The one input we CAN see moved against the carried price: USD/JPY is 159.30 this morning against a 159.22 settle on Aug 25, still just under the level that provoked joint intervention. A weaker yen argues for the BoJ moving. Crude around $85.71 argues against. With no fresh price we are not revising our 84 fair value in either direction. Tokyo CPI is due tomorrow and the decision is Sep 18.
Fed Decision in September - 25bp increase
32.5%→32.5%0.0pp
CARRIED from Aug 26 - Polymarket unreachable all session. Two anchors now unusable at once: the Polymarket price is carried, and CME FedWatch has been unverifiable at source for a seventh consecutive session, leaving a stale 31.6% dated Aug 20. We decline this leg for an eighth time. Yesterday's flagged inconsistency - the September leg falling while our full-year contract did not - cannot be re-examined today because neither price is current.
US 10Y Treasury
4.639%→4.664%+0.025pp
Aug 26 settle, verified. The only rate in this letter we can stand behind today. It rose despite crude falling again, which is consistent with the inflation print: headline 3.7% and core 3.3% both came in above consensus, and 30Y at 5.186% moved up with it. Two sessions of a rising 10-year is also what disarmed the trim rule we wrote on Aug 26, one day after writing it.
Screening Table
| # | Market | Expiry | Market Price | Fair Value | Gap (pp) | Direction | Volume | Confidence |
|---|---|---|---|---|---|---|---|---|
| 1 | 10Y Touches 4.8% Before 2027 (pos-017) | Dec 31 | 65.5% | 61% | -4.5pp | HOLD $25 YES - model says 65, our own rule freezes it at 61; price carried, gates unevaluable | $$60K on the leg | 3/10 |
| 2 | BoJ September 25bp increase | Sep 18 | 86.0% | 84% | -2.0pp | NO POSITION - yen weaker, crude weaker, no fresh price; FV unchanged in both directions | $$119K on the leg | 2/10 |
| 3 | ECB September 25bp increase | Sep 10 | 96.75% | 96% | -0.75pp | NO POSITION - 3.25 points of upside against the full stake; carried price, untestable thesis | $$92.6K on the leg | 3/10 |
| 4 | Fed Decision in September - 25bp increase | Sep 16 | 32.5% | 31.6% | -0.9pp | NO POSITION - carried price AND a stale Aug 20 anchor; declined for an 8th session | $$15.5M on the leg | 1/10 |
| 5 | 10Y Touches 5.00% Before 2027 | Dec 31 | 17.0% | 34.0% | +17.0pp | REFUSED for a 3rd session - the gap WIDENED only because our barrier narrowed; the market leg was never re-observed | $$82.7K on the leg | 1/10 |
Top 5 Opportunities
1
10Y Treasury Touches 4.80% Before 2027 - the arithmetic says 65, our rule says 61, and the rule wins — YES
↑ BUY YES-4.5pp
Market price
65.5%
Fair value
61%
Gap: -4.5pp
The US 10-year settled at 4.664% on Wednesday, up 2.5 basis points and up 2.5 from Tuesday's 4.639%. The barrier to 4.80% is now 13.6 basis points with 90 sessions remaining, giving a horizon sigma of 37.95bp. The model returns 72.0% raw and 65.1 after the buyback haircut we have applied since entry. That is a four-point upgrade on yesterday's 61. We are not taking it, and the reason is a rule we wrote two days ago for exactly this situation. On Aug 25 we said there would be no fair-value increase on this position until the 5.00% consistency check reproduces. Today it does not. The model gives 34.0 on the 5.00% leg against a 17.0c market - a 17pp gap, wider than yesterday's 14.1pp. And that widening is itself an artefact: it happened because the 10-year rose and both barriers narrowed, while the market leg was never re-observed, because Polymarket has been down all session. So we have a favourable revision blocked by a rule, tested against a check we cannot run properly, using a comparison price that is a day old. In that situation the conservative reading is the only defensible one. FV stays 61 and our published book understates the position by four points on our own numbers. The trim rule armed yesterday was also checked: it needs two consecutive settles below 4.60%, and with 4.639% then 4.664% it did not fire. It stays armed.
▵ Bull case
- Barrier 13.6bp over 90 sessions gives 72% raw before any haircut - the narrowest barrier this position has had since entry.
- Our own arithmetic says 65 and we are publishing 61, so the stated mark is conservative by construction.
- Hot core PCE at 3.3% and GDPNow at 4.6% are the combination that pushes long yields up.
- The Treasury buyback window expires Nov 4 while the barrier runs to Dec 31.
▿ Bear case
- The 5.00% check has failed three sessions running and the model behind every number here remains one we have publicly said we cannot reconcile.
- Every market price in this section is carried. We cannot verify the 65.5, the 17.0, or whether the 66c take-profit bid was touched.
- Crude near $85.71 continues to take war premium out of the long end.
- Flat real consumer spending is the classic precursor to lower yields, and it showed up in the same release as the hot inflation print.
2
Bank of Japan September 25bp increase - the yen moved against the carried price and we are still not revising — YES
↑ BUY YES-2.0pp
Market price
86.0%
Fair value
84%
Gap: -2.0pp
Carried at 86.0 from Aug 26 and unverifiable. What we can see is USD/JPY at 159.30 this morning against a 159.22 settle on Aug 25, which keeps the pair just under the level that provoked the joint intervention earlier this month. That drift argues for the BoJ moving: yen weakness is the most durable component of Japan's imported-inflation problem, and it has not improved. Pulling the other way, crude at roughly $85.71 is around 9% below Friday, and for a country that imports essentially all of its oil that is a direct and continuing improvement in the terms of trade. Those two arguments have not changed in relative weight since yesterday, and with no fresh price to test them against we are leaving fair value at 84 rather than manufacturing a revision out of a currency tick. We continue to hold no position. At a carried 86c the contract pays about 1.16-to-1 into a meeting three weeks out, and our fair value sits two points below the market. The reason we keep writing about it is that it remains the cleanest available read on whether the Hormuz de-escalation is being priced as durable. Tokyo CPI tomorrow is the next real input.
▵ Bull case
- USD/JPY at 159.30 is drifting back toward intervention territory, and yen weakness is the BoJ's most persistent argument for normalising.
- JGB 10-year at 2.8950% - carried for two sessions now - sits at a 1996 high.
- The Sep 18 meeting is scheduled and the direction has been telegraphed repeatedly.
- US core PCE at 3.3% keeps the rate differential wide, which pressures the yen further.
▿ Bear case
- Crude around 9% below Friday is a direct terms-of-trade gain for Japan and erodes the imported-inflation case.
- At 86c there are 14 points of upside against 86 of downside.
- The price is carried, so we do not know whether the market has already moved on any of this.
- Our own fair value of 84 is below the market, so agreeing with the direction still means overpaying.
3
ECB September 25bp increase - yesterday's reading survives the day only because nothing could test it — YES
↑ BUY YES-0.75pp
Market price
96.75%
Fair value
96%
Gap: -0.75pp
Carried at 96.75 from Aug 26. Yesterday we offered a reading: that this contract has stopped pricing whether the economics justify a September hike and started pricing whether a telegraphed decision gets executed, which would explain why it ROSE on a 10% fall in imported energy. We flagged that as a reading rather than a finding, and today we have to report that it remains exactly that. With no fresh price, nothing about it was tested. The US inflation print is weakly consistent with it - hot core with no obvious euro-area read-across should not move a contract that has stopped trading on inflation - but that is a very thin form of confirmation and we are not going to dress it up as more. There is nothing here for us either way. Even taking the reading at face value, 3.25 points of upside against the entire stake fails our sizing rules at any confidence level. Euro-area flash HICP on Aug 31 is the first thing that could genuinely test the thesis, and the decision follows on Sep 10.
▵ Bull case
- Bid 96.7 / ask 96.8 when last observed - a tight two-sided market rather than a thin constructed mid.
- The decision is in fourteen days and has been telegraphed extensively.
- Cheaper energy supports euro-area growth even as it lowers headline inflation, removing the growth objection to hiking.
- Bund near 3.2550% - carried, now two sessions stale - is consistent with a curve that has absorbed a September move.
▿ Bear case
- 3.25 points of upside against the full stake fails every sizing rule we run.
- Our explanation for the price is unverified and stayed unverified today.
- Both the price and the Bund level supporting the view are stale, the Bund by two sessions.
- Fourteen days is long enough for a formal Hormuz announcement and a flash HICP print to change the picture.
4
Fed Decision in September - two broken anchors at once, so we decline for an eighth session — YES
↑ BUY YES-0.9pp
Market price
32.5%
Fair value
31.6%
Gap: -0.9pp
This is the deepest market on the board and we still will not touch it, for a reason that got worse today rather than better. Our fair value here has always been a CME FedWatch implied probability, and FedWatch has now been unverifiable at source for seven consecutive sessions. The last figure we could actually stand behind is 31.6%, dated Aug 20. As of today the market price is also carried, at 32.5c from Aug 26. So we are being asked to compare a seven-day-old anchor against a one-day-old price, and to call the difference an edge. It is not an edge, it is two errors of unknown size. Today's data would ordinarily be exactly the input that repriced this leg: core PCE above consensus, GDPNow at 4.6%, durable goods beating. All of that argues the September hike probability should have risen. We simply do not know whether it did, and we are not going to construct a fair value from our own reasoning and then congratulate ourselves on the gap to a price we cannot see. Eighth consecutive refusal. We will re-price this leg on the session where both FedWatch and Polymarket are verifiable, and not before.
▵ Bull case
- $15.5M of two-sided volume makes this, when observable, the most reliable price on the board.
- Core PCE above consensus and GDPNow at 4.6% both argue the September hike probability should be rising.
- Three FOMC members dissented in July in favour of hiking and have now been handed a hot print.
- Warsh speaks tomorrow, which is a genuine catalyst before the Sep 15-16 meeting.
▿ Bear case
- Our fair value is seven days old and the market price is one day old. Neither is a basis for risking money.
- Crude roughly 9% below Friday reduces the near-term inflation impulse into the meeting window.
- Warsh's stated aversion to forward guidance makes tomorrow unlikely to resolve anything cleanly.
- A 0.9pp gap is far smaller than the uncertainty attached to either input.
5
10Y Treasury Touches 5.00% Before 2027 - the gap grew to 17pp and that is exactly why we still will not take it — YES
↑ BUY YES+17.0pp
Market price
17.0%
Fair value
34.0%
Gap: +17.0pp
On the fresh 10-year settle the barrier to 5.00% is 33.6 basis points, and over 90 sessions with horizon sigma 37.95bp the model returns 37.6% raw and 34.0 after the standard haircut. Against a carried 17.0c market that is a 17pp gap - up from 14.1pp yesterday and 22pp the day before. A widening gap on a screening table is normally a reason to look harder. Here it is the opposite, and the reason is worth stating precisely. The gap widened because the 10-year rose 2.5 basis points, which narrowed the barrier and lifted the model's output. The market leg did not move, because we could not observe it - Polymarket has been down all session. So the entire improvement in this 'opportunity' is one side of a comparison updating while the other side stands still. Treating that as a bigger edge would be a straightforward way to talk ourselves into a trade using our own stale data. The structural objection also stands, unchanged and now three sessions old. Fixed-volatility barrier models over-price far touches by construction, because they assume the world stays as volatile at 5.00% as it is at 4.66%, when a sustained grind toward a round number the Treasury has visibly defended would summon exactly the response that stops it. And the portfolio objection from Wednesday has not expired: this would be a fifth correlated expression of a long-end trade in a book that just watched one oil headline move four of its five positions together. Refused.
▵ Bull case
- The nominal gap is 17pp and has exceeded our 10pp entry bar on three consecutive sessions.
- Gilts above 5% - carried - prove a G7 sovereign can trade there.
- The buyback window expires Nov 4 while the barrier runs to Dec 31, leaving a genuinely unprotected stretch.
- At 17c the payout is close to 5-to-1, so only a modest true edge would be needed on a small stake.
▿ Bear case
- The gap widened only because our own input updated and the market's did not. That is an artefact, not an opportunity.
- The model has failed its own consistency check three sessions running.
- Three official interventions this month demonstrate the reaction function that would stop the move, and none of it is in the model.
- It would be a fifth correlated position in a book whose correlation cost was demonstrated two days ago.