Daily Macro US
Treasury buybacks
$2bn → $4bn
Long-end liquidity ops at least doubled, Sep 9–Nov 4, 10-20y and 20-30y sectors
10Y Treasury
~4.64%
−6bp Wednesday to 4.647%; 16bp from pos-017's touch, but a buyer of size now stands in the way
30Y Treasury
5.196%
−9bp Wednesday, one day after 5.34% — a nineteen-year high
Brent crude
<$92
+5% over four sessions, Hormuz still shut, 8 vessels attacked this month — and yields fell anyway
Fed Sep hike
29c
CME FedWatch 34.6%; gap 5.6pp, under the bar. Warsh at Jackson Hole Aug 28
Book
+$87.91
5 open, $332 staked; +$611.08 realized, 11/14
The Treasury Department took our best position's engine away, and we are cutting our own fair value by 14 points because of it. On Wednesday the Treasury announced it is at least doubling the maximum size of its liquidity-support buyback operations in longer-dated nominal coupons, from $2bn to at least $4bn per operation, covering the 10-to-20-year and 20-to-30-year sectors, effective September 9 and running through November 4. The long end responded immediately: the 10-year shed about 6bp to 4.647% and the 30-year gave up 9bp to 5.196%, one day after the 30-year had printed 5.34%, a nineteen-year high. This morning the 10-year is around 4.64%. That matters more to this book than any data release, because pos-017 is a barrier bet that the 10-year touches 4.80% before 2027, and there is now a buyer of size standing in the way of exactly that. Note what did NOT happen: Brent traded below $92 after gaining more than 5% across four sessions, the Strait of Hormuz is still shut, eight vessels have been attacked in transit this month, and on Wednesday President Trump announced a fresh package aimed at crushing Iran's economy — and the 10-year fell anyway. For six weeks our entire thesis on pos-017 has been that oil drives the term premium. This week oil went up and the term premium went down. When your mechanism stops working you write it down, so: fair value on pos-017 comes down from 74 to 60. The barrier arithmetic alone still says 68% — 16bp to go across roughly 93 sessions at about 4bp of daily vol — but a haircut for a policy actor actively suppressing the long end through November takes us to 60. The market is 64 mid, 58 bid / 70 ask. That means for the first time since we bought it at 16.5c, we mark this position ABOVE our own fair value. We are not selling into a 58c bid that is below fair value, but we are lowering the published take-profit gate from a 72c bid to a 66c bid, because a gate should sit above fair value and 72 no longer does. DISCLOSURE FIRST, as always: there was no letter Friday Aug 14, Monday Aug 17, Tuesday Aug 18 or Wednesday Aug 19. Four consecutive missed sessions, and they contained the 30-year's nineteen-year high and the Treasury's response to it. Nothing in this note was decided during the blackout, but readers who rely on this letter to track a live book were unserved for a week, and the two largest fair-value errors it now corrects both developed inside the gap. That is the second attendance failure in three weeks and it is the biggest problem this project has. The rest of the tape: pos-011 vindicated our contrarian cross-market read, rising from 22.65c to 29.6c — we said the exactly-one-hike bracket reconciled near 30 and it is now 29.6, so the call was right and the edge is now gone. pos-010 fell to 50c and sits just two points above its published 48c floor. July retail sales came in at -0.6% on the month, July PPI final demand was unchanged on the month at +4.7% year-over-year, and initial claims for the Aug 8 week were 209K against a 202K consensus with the four-week average still at 199K. This week's claims print lands at 8:30 ET today and is not out at the time of writing. Polymarket has the September hike at 29c against CME FedWatch's 34.6% — a 5.6pp gap, under our bar, and a fifth correlated hawkish ticket besides. Kevin Warsh gives his first Jackson Hole keynote as Chair on Friday Aug 28, nineteen days before the September 16 FOMC. No trades today. Book: 5 open, $332 staked, +$87.91 unrealized, +$611.08 realized, 11/14.
Today's Market Moves
10Y Touches 4.8% (pos-017)
68.0%→64.0%-4.0pp
FV cut 74 → 60, our largest single correction on this position. Cause is not the tape, it is a new actor: Treasury at least doubled long-end buyback operations to $4bn, effective Sep 9 through Nov 4, and the 10-year fell 6bp to 4.647% on the announcement while Brent was still climbing. Our published mechanism — oil drives the term premium — failed a clean test this week. Barrier maths on 16bp over ~93 sessions at 4bp daily vol gives 68%; the buyback haircut takes it to 60. At 64 mid we now mark above our own FV. Take-profit gate lowered from a 72c bid to a 66c bid.
Fed Funds End 2026 = 4.0% (pos-011)
22.65%→29.6%+6.95pp
Called and confirmed. On Aug 13 we said the cross-market arithmetic reconciled exactly-one-hike-and-hold near 30 while the bracket printed 22.65, and refused to call it broken. It is 29.6 today, a 7-point recovery, and the position is back to −$3.24 from −$8.4. The honest consequence is that the edge is now zero: FV 30 against 29.6. We hold to resolution and add nothing; efficiency review armed at 40c.
Fed Rate Hike in 2026 (pos-010)
54.5%→50.0%-4.5pp
Down to a round 50c and now only two points above the 48c floor we published in June. FV cut 58 → 55: July retail sales at −0.6% and claims drifting to 209K soften the hike case, oil argues the other way, and the internal cross-check holds (hike-by-October 41c versus 50c for the full year leaves ~9pp of December-only hike, which is coherent). If the runner trades below 48c we exit it, in public, on the same day.
Fed Sep Hike (context)
32.5%→29.0%-3.5pp
Third consecutive session of drift lower on Polymarket while CME FedWatch reads 34.6% for a hike at the Sep 15-16 meeting. That is a 5.6pp gap — the same shape of trade we refused on Aug 6 and Aug 13, and refusing it has been right both times. Under the bar, and it would be a fifth correlated hawkish Fed ticket. Warsh's Jackson Hole keynote on Aug 28 is the catalyst, and he has said publicly he will not be constrained by market pricing.
Zero Fed Cuts 2026 (pos-013)
85.95%→85.3%-0.65pp
Flat, correctly. Neither the softer consumer nor the Treasury's long-end intervention is an argument for the Fed cutting in 2026 — the buyback is a fiscal-financing tool aimed at the long end, not a policy signal about the front end. FV stays 88; efficiency trim armed at 90c and not touched.
US unemployment >= 5.0% (pos-004)
10.1%→9.9%-0.2pp
Still drifting our way. July unemployment was 4.1%, down from 4.2% in June, and claims at 209K with a 199K four-week average are not a recession signal. Our NO is worth 90.1c against a 75c entry, +$20.13. The published early-exit review triggers if YES trades below 8c; at 9.9c it is close enough to name again, and the decision at 8c will be a review, not an automatic sale.
Screening Table
| # | Market | Expiry | Market Price | Fair Value | Gap (pp) | Direction | Volume | Confidence |
|---|---|---|---|---|---|---|---|---|
| 1 | 10Y Touches 4.8% Before 2027 | Dec 31 | 64.0% | 60% | -4pp | BUY YES | $$286K total | 6/10 |
| 2 | Fed Rate Hike in 2026 | Dec 9 | 50.0% | 55% | +5pp | HOLD $100 runner — no add; floor at 48c is two points away | $$7.7M total | 6/10 |
| 3 | Fed Sep Hike | Sep 16 | 29.0% | 34.6% | +5.6pp | NO TRADE — under the 10pp bar; fifth correlated hawkish ticket | $$869K total | 4/10 |
| 4 | Fed Funds End 2026 = 4.0% | Dec 8 | 29.6% | 30% | +0.4pp | HOLD $25 — thesis confirmed, edge fully closed | $$1.37M total | 5/10 |
| 5 | Zero Fed Cuts in 2026 | Dec 30 | 85.3% | 88% | +2.7pp | HOLD $82 — trim armed at 90c | $$7.3M total | 7/10 |
Market vs Fundamentals
Market Price (red) vs Estimated Fair Value (green) — %
Top 5 Opportunities
1
10Y Treasury Touches 4.8% Before 2027 — YES
↑ BUY YES-4pp
Market price
64.0%
Fair value
60%
Gap: -4pp
This is the entry that costs us the most to write. For six weeks the published thesis on this position has been a single mechanism: the Iran conflict keeps oil elevated, elevated oil rebuilds the inflation risk premium, and the premium shows up in the long end. On Aug 10 we corrected fair value up from 55 to 65 on that logic, and on Aug 13 up again to 74 when Brent went to $90 and the tape agreed within three sessions. This week the mechanism was tested cleanly and failed. Brent gained more than 5% across four sessions and traded below $92, Hormuz remains shut with eight vessels attacked in transit this month, Trump announced fresh measures against Iran's economy on Wednesday — and the 10-year FELL, 6bp, to 4.647%. What moved it was fiscal, not geopolitical: the Treasury announced it is at least doubling the maximum size of its long-end liquidity-support buybacks, from $2bn to at least $4bn per operation, across the 10-20y and 20-30y sectors, effective September 9 and running to November 4. That is a policy actor with an unlimited balance sheet standing on the exact part of the curve our bet needs to move. The barrier arithmetic is unchanged and still friendly: 16bp to travel, roughly 93 sessions left, about 4bp of daily vol, which gives a 68% touch probability. But the haircut is now much larger than the 5% we applied for a dovish front end, because for nearly two of the remaining four and a half months the Treasury is actively bidding the long end. We take fair value to 60. The position is 64 mid, 58 bid, 70 ask, so we are now marked above our own fair value — and we are saying so rather than quietly leaving the number at 74. We do not sell into a 58c bid that is below fair value; we lower the take-profit gate from a 72c bid to a 66c bid, because a gate that sits below fair value is not a gate, it is a wish.
▵ Bull case
- 16bp with ~93 sessions left still gives 68% on the raw barrier maths
- The buyback window ends November 4; December is unprotected
- Highest print all year was 4.75% on July 31 — the barrier has been approached once already
▿ Bear case
- Our published mechanism failed a clean test: oil up, yields down
- Treasury is bidding the exact sectors we need to sell off, through November 4
- Fifth fair-value revision in six weeks on this position — that is instability, and we own it
- 58 bid / 70 ask means the 64 mid we carry is notional
2
Fed Funds Rate End 2026 = 4.0% (called and confirmed) — YES
↑ BUY YES+0.4pp
Market price
29.6%
Fair value
30%
Gap: +0.4pp
Seven sessions ago this was the worst position in the book at −$8.4 and 22.65c, and the easy thing to write would have been that the thesis was broken. Instead we published arithmetic: the liquid hike-in-2026 market was at 54.5, the brackets at or above 4.25% summed to about 20, roughly 4 points belonged to a hike that gets reversed, and exactly-one-hike-and-hold therefore reconciled near 30 — so the bracket was cheap, not broken. It trades at 29.6 today. We publish this the same way we publish the misses: the number we wrote down in advance was 30 and the market walked to it, which is the cleanest confirmation this process produces. The uncomfortable half is that being right closes the trade. At 29.6 against a fair value of 30 there is 0.4pp of edge, which is nothing, and we will not manufacture a reason to add to a fifth correlated Fed ticket in a market doing under $100 a day. So we hold the original $25 to resolution and arm an efficiency review at 40c — if it gets there, the capital is worth more somewhere less correlated.
▵ Bull case
- The pre-published cross-market reconciliation at 30 was hit within seven sessions
- Nothing this week argued against exactly one hike: retail sales soft, oil firm
- Small ticket — $25 of a $332 book, cheap to hold to resolution
▿ Bear case
- Edge is now 0.4pp; the trade is over even though the position is not
- Under $100/day of volume makes any exit expensive
- Still a Fed ticket in a book that is mostly Fed tickets
3
Fed Rate Hike in 2026 (runner) — YES
↑ BUY YES+5pp
Market price
50.0%
Fair value
55%
Gap: +5pp
The half we banked at 66.5c in July looks better every week. The runner is at exactly 50c, which is 4.5 points lower than a week ago and only two points above the 48c floor we published on June 10. We are naming that in advance so nobody is surprised: if this trades below 48c we sell the runner and print the loss in the next letter, on the day. Fair value comes down from 58 to 55. The case for the hike is intact — oil is still the dominant inflation input, Hormuz is shut, and three FOMC members dissented in favour of a hike at the July meeting — but the case against has strengthened on the consumer side: July retail sales fell 0.6% on the month, claims rose to 209K against a 202K consensus, and the July PPI final-demand index was unchanged on the month. The internal cross-check still reconciles: hike-by-October trades at 41c against 50c for the full year, leaving about 9 points of December-only hike, which is a sensible shape rather than an arbitrage. At 50c against a fair value of 55 the gap is 5pp, under the bar, so no add.
▵ Bull case
- Three July dissents in favour of a hike; the committee is genuinely split
- Oil above $90 keeps the inflation input live regardless of the term premium
- Internal term structure (41c October vs 50c full-year) is coherent, not broken
▿ Bear case
- Two points from a published floor we will honour
- Consumer is cracking: retail sales −0.6%, claims 209K, PPI flat
- Warsh has said he will not be constrained by market pricing — cuts both ways
4
Fed September Hike (screened — refused for a third time) — YES
↑ BUY YES+5.6pp
Market price
29.0%
Fair value
34.6%
Gap: +5.6pp
This is the third consecutive letter in which this market screens and gets refused, and the arithmetic keeps landing in the same place. Polymarket prices the September meeting hike at 29c. CME FedWatch, as of Wednesday, priced roughly 34.6% for a hike against 65.4% for a hold. That is 5.6pp of gap — real, verified at two independent sources in this session, and still comfortably under the 10pp bar we publish. On Aug 6 we refused a version of this trade at 44.5c because we could not verify the quoted gap at source, and that refusal has since been worth about 15 points. On Aug 13 we refused it at 32.5c against a 38% FedWatch read, and that has been worth another 3.5. Refusing has now been the right decision twice in a row on the same ticket, which is worth stating plainly given that the last letter led with a refusal that cost us $41.67. The bar does not know which refusals will look clever afterwards; that is the entire point of it. The second reason stands independently: the book already holds four positions whose payoff depends on the Fed not cutting and probably hiking. A fifth would not be a trade, it would be leverage on an existing view.
▵ Bull case
- A genuine 5.6pp gap between Polymarket and CME, verified at both sources today
- Warsh's Jackson Hole debut on Aug 28 is a real repricing catalyst before the meeting
- Deep market at $869K of volume — sizeable if we ever did want it
▿ Bear case
- 5.6pp is not 10pp; the bar is the bar
- Fifth correlated hawkish-Fed ticket — the correlation cap forbids it outright
- Softening consumer data cuts against the hike case, not for it
5
Zero Fed Rate Cuts in 2026 — YES
↑ BUY YES+2.7pp
Market price
85.3%
Fair value
88%
Gap: +2.7pp
The quietest position in the book and the one that has needed the least explaining, at 85.3c against an 85.95 mark a week ago and +$8.14 on an $82 stake. It is worth spelling out why this week's news does not touch it, because the temptation is to read the Treasury buyback as a monetary event. It is not. Doubling long-end liquidity-support operations is a debt-management action taken by the Treasury on the 10-to-30-year sectors; it says nothing about the front end, which is where cuts live. If anything, a Treasury Secretary reaching for buybacks because the 30-year hit a nineteen-year high is evidence of a term-premium problem the Fed has explicitly declined to solve by easing. Meanwhile the front-end data gives no cut case at all: unemployment fell to 4.1% in July, claims are running near 200K on the four-week average, and the FOMC has three members who dissented toward a hike. Fair value stays 88. The efficiency trim stays armed at 90c — at that price the remaining 12 points of upside are not worth $82 of capital in a book with only five positions.
▵ Bull case
- Unemployment 4.1%, claims four-week average 199K — no cut case in the labour data
- Three July dissents were toward a hike, not a cut
- The Treasury buyback is a debt-management action, not an easing signal
▿ Bear case
- 85.3c leaves 14.7 points of upside for $82 of capital — poor efficiency
- A genuine growth shock in Q4 would reprice this fast
- Largest single stake in the book on the least dramatic idea