Daily Macro US
pos-017 fair value
61 (FROZEN)
Arithmetic says 66.8 on a 12.6bp barrier. The Aug 25 standing rule blocks the increase and the freeze now understates the book by 5.8 points, up from four yesterday. Kept.
5.00% consistency check
17.6pp gap, BOTH LEGS LIVE
Model 35.1 vs a fresh 17.5c market. Yesterday's 17pp was partly an artefact of a carried market leg; today nothing is carried and the gap is wider. The excuse is gone.
10Y Treasury
4.674% (Aug 27, PROVISIONAL)
Up ~1.0bp from the 4.664% Aug 26 settle. FRED/H.15 still ends Aug 26 at 4.66 and publishes Aug 27 today, so the third decimal is vendor-sourced and will be re-checked.
Polymarket feed
LIVE
Restored after Wednesday's full-session outage. Every mark in this letter is fresh and every price-triggered gate is checkable again for the first time in two days.
Cleveland Aug core PCE nowcast
3.40% YoY
Re-verified at source this morning and UNCHANGED, stamp advanced to 08/27. Headline 3.79%, core CPI 2.38%, Q3 core PCE 3.00% annualised.
Today 10:00 ET (PREVIEW)
Warsh keynote + BLS benchmark
Both land six hours after we publish. Written as previews, not results. The pos-013 trim at 90c sits 1.3c away and may fire while we are not watching.
Two things changed about this letter today and you should know both before reading a single number. The first is that the Polymarket feed came back. Wednesday we published a whole book of carried prices and could not evaluate a single price-triggered gate; today every mark is live, the +$104.95 open figure is real, and the gates are checkable again. The second is that we now publish at 04:06 ET rather than 08:40 ET. That is a deliberate change and it has a cost we are going to state at the top rather than bury: Warsh's first Jackson Hole keynote and the BLS preliminary benchmark payroll revision both land at 10:00 ET, six hours after this goes out. Everything we write about them today is a PREVIEW. Nothing below is a result. And the restored feed did something uncomfortable to yesterday's story. On Wednesday the 5.00% consistency check failed by 17 points and we partly excused the widening as an artefact - the 10-year had risen so our model lifted, while the market leg was carried from Tuesday and never re-observed. Today both legs are fresh. The 5.00% market prints 17.5c, the model gives 35.1, and the gap is WIDER at 17.6 points. There is no artefact left to blame. That is a cleaner failure than yesterday's, and a cleaner failure makes the Aug 25 standing rule bind harder rather than softer. So pos-017 fair value stays frozen at 61 while the arithmetic says 66.8, and the freeze now costs us 5.8 points of published book value instead of four. We are keeping it. A rule you abandon the third time it hurts was never a rule. Two smaller admissions in the same spirit. The 4.674% Aug 27 10-year settle underneath all that arithmetic is provisional: FRED and H.15 still end at Aug 26 and publish yesterday's figure later today, so our third decimal comes from a vendor series rather than the primary source. And while re-running the barrier model we found our own session count disagrees with a direct calendar count - 89 against 84 - which is worth 0.7 of a point today and is flagged now, before it is worth more.
Today's Market Moves
10Y Touches 4.8% Before 2027 (pos-017)
65.5%→64.5%-1.0pp
LIVE, first fresh mark in two sessions. The model went the other way: 10Y 4.674% narrows the barrier to 12.6bp and lifts the arithmetic to 66.8. FV stays 61 because the consistency check failed again, this time with nothing carried to explain it away.
10Y Touches 5.00% Before 2027 (check leg)
17.0%→17.5%+0.5pp
THE NUMBER THAT MATTERS TODAY. Carried at 17.0 on Wednesday, live at 17.5 now. Our model says 35.1 on the same barrier method that produces the 4.8% figure we trade. A 17.6pp disagreement on fresh data is a model we do not trust far from the money, which is exactly what the standing rule was written for.
Zero Fed Rate Cuts in 2026 (pos-013)
86.45%→88.7%+2.25pp
Best mark this position has held. Claims 203k against ~208k expected, the Cleveland core nowcast holding 3.40%, and Brent settling UP 0.77% all point one way. FV 88 -> 89. The 90c trim is now 1.3c away and we publish before the US session.
Fed Rate Hike in 2026 (pos-010)
56.5%→57.5%+1.0pp
FV 54 -> 55, continuing Wednesday's partial restoration, because the bear input partly reversed: Brent settled $88.52 on Aug 27, up 0.77% from $87.84, so the crude collapse we cut on has stopped collapsing. Raised into an event we cannot see.
US Unemployment 2026 >= 5.0% (pos-004)
12.5%→11.2%-1.3pp
Moved our way on a live mark. Claims 203k with the insured rate steady at 1.2% is not an economy shedding the ~1.4m additional unemployed this contract needs. The benchmark revision at 10:00 ET is the only real event this position has left.
Fed Rate End 2026 = 4.0% (pos-011)
31.05%→30.65%-0.4pp
FV 30 -> 31 derived mechanically from pos-010, as every move in this leg has been. Full ladder verified fresh for the first time in three sessions: 3.5% 7.55 / 3.75% 33.7 / 4.0% 30.65 / 4.25% 13.45 / >=4.5% 4.4.
Screening Table
| # | Market | Expiry | Market Price | Fair Value | Gap (pp) | Direction | Volume | Confidence |
|---|---|---|---|---|---|---|---|---|
| 1 | 10Y Touches 4.8% Before 2027 | Dec 31 | 64.5% | 61% | -3.5pp | HOLD $25 YES - arithmetic says 66.8, our own rule freezes it at 61; the check failed again on live data | $$60K on the leg | 3/10 |
| 2 | Zero Fed Rate Cuts in 2026 | Dec 31 | 88.7% | 89% | +0.3pp | HOLD $82 YES - FV 88 -> 89 on claims, a steady core nowcast and firmer crude; trim armed 1.3c away | $$7.6M | 4/10 |
| 3 | US Unemployment 2026 >= 5.0% | Dec 31 | 11.2% | 8% | +3.2pp | HOLD $100 NO - the benchmark payroll revision lands at 10:00 ET, after we publish | $$124K on the leg | 4/10 |
| 4 | Fed Rate Hike in 2026 | Dec 31 | 57.5% | 55% | -2.5pp | HOLD runner $100 YES - FV 54 -> 55 as the crude bear case partly reversed; Warsh is the event | $$8.0M | 3/10 |
| 5 | Fed Rate End 2026 = 4.0% | Dec 31 | 30.65% | 31% | +0.35pp | HOLD $25 YES - FV 30 -> 31 tracking pos-010; too small to trade around | $$1.37M on the leg | 3/10 |
Market vs Fundamentals
Market Price (red) vs Estimated Fair Value (green) — %
Top 5 Opportunities
1
10Y Treasury Touches 4.80% Before 2027 - the excuse is gone and the freeze got more expensive — YES
↑ BUY YES-3.5pp
Market price
64.5%
Fair value
61%
Gap: -3.5pp
Yesterday we froze fair value at 61 against an arithmetic 65 and offered a partial defence of the failing consistency check: the gap had widened to 17 points mostly because the 10-year rose, which lifted our model, while the 5.00% market leg was carried from Tuesday and never re-observed. An unresolvable check, we said, is not a passed check. Today the feed is back and the check is fully resolvable, so that defence is now testable - and it fails. The 5.00% leg prints 17.5c live. The model, on the same barrier method that generates the 4.8% number we actually trade, says 35.1. The gap is 17.6 points, wider than yesterday, on two fresh observations with nothing carried. That is the cleanest version of the failure we have had, and it says something specific: a fixed-volatility barrier model systematically over-prices far touches, so the 4.8% output it hands us is probably also too high, just by less. Meanwhile the arithmetic on our own leg improved. The 10-year settled 4.674% on Thursday, narrowing the barrier to 12.6 basis points and lifting the raw model to 73.8%, or 66.8 after the buyback haircut. So the freeze now understates this position by 5.8 points rather than four. We are keeping it, and we would rather publish a book that is wrong in a direction we chose than one that is right by abandoning a rule the third time it costs us.
▵ Bull case
- Barrier is only 12.6bp with 89 sessions left. On any reasonable volatility that is a high-probability touch, and the model says 66.8 after haircut against a 64.5 market.
- The 10-year has risen two consecutive sessions, 4.639% to 4.664% to 4.674%, moving toward the barrier rather than away.
- Global long ends are pulling the same way: the Bund is at a 15-year high at 3.2648%, Ireland's 10-year hit a 12.5-year high, and the JGB is up 3.7bp to 2.9320%.
- The buyback window expires Nov 4 while the barrier runs to Dec 31, leaving nearly two months of unsupported duration at the end.
▿ Bear case
- Our own consistency check fails by 17.6pp on fresh data on both legs, which is direct evidence the model over-prices touches and that even 66.8 is too generous.
- The market has held 64-66c for days and is a better estimator than a fixed-vol model we have twice failed to reproduce.
- The Aug 27 settle of 4.674% is provisional - FRED/H.15 still ends Aug 26 and publishes today, so the input to all of this arithmetic is vendor-sourced.
- Our house session count (89) disagrees with a direct calendar count excluding bond holidays (84). Immaterial today at 0.7 of a point, but it is an unexplained error in our own method.
2
Zero Fed Rate Cuts in 2026 - our best mark, and a trim rule that may fire while we are not looking — YES
↑ BUY YES+0.3pp
Market price
88.7%
Fair value
89%
Gap: +0.3pp
88.7c is the highest this position has traded since we entered at 77.6c, and the move is backed by data rather than drift. Initial claims came in at 203,000 against roughly 208,000 expected, down 4,000 on a week whose prior level was revised up to 207,000, with the insured unemployment rate steady at 1.2%. The Cleveland August nowcast, re-verified at source this morning, is unchanged at 3.40% core PCE and 3.79% headline. And Brent settled $88.52 on Thursday, up 0.77% from Wednesday's $87.84, which takes back part of the crude-driven disinflation we credited on Tuesday when we cut this fair value. A committee does not cut into 3.4% core inflation with claims printing 203k. FV goes 88 to 89. Now the part that costs us something. The efficiency trim on this position is armed at 90c. The market is 1.3c away. From today this letter is written at 04:06 ET instead of 08:40 ET, which means that if this leg trades through 90c during the reaction to Warsh at 10:00 ET, we will not observe it until the next run. We are stating that in advance rather than discovering it afterwards. The rule is not weakened by our not watching it: it stands exactly as written, it gets evaluated against today's actual prints at the next observation, and if it traded through while we were away we will say so and act then, rather than quietly re-dating the trigger to whenever we happened to look.
▵ Bull case
- Claims at 203k against ~208k expected, insured rate steady at 1.2% - no labour-market deterioration to cut into.
- Cleveland August core PCE nowcast holding 3.40% YoY, headline 3.79%, re-verified at source and unchanged.
- July core PCE printed 3.3% and headline 3.7%, both above the consensus we had published.
- Brent settled up 0.77% on Thursday, removing part of the autumn disinflation impulse that argued for cuts.
▿ Bear case
- Inflation-adjusted consumer spending was FLAT in July after +0.4% in June. This remains the actual mechanism by which the position loses and it has not been retracted.
- At 88.7c there are 11.3 points of upside against 88.7 of downside; the risk-reward is now poor even if the thesis is right.
- Warsh speaks at 10:00 ET and we cannot see it. A chair with an explicit bias against forward guidance can move this leg in either direction.
- Our FV of 89 is barely above the market, so we are holding mostly on inertia and a trim rule rather than on measured edge.
3
US Unemployment 2026 >= 5.0% - the only event this position has left lands after we publish — NO
↑ BUY YES+3.2pp
Market price
11.2%
Fair value
8%
Gap: +3.2pp
The YES fell to 11.2c on a live mark, down 1.3 from the 12.5 we carried, so the NO we hold is worth 88.8c against a 75c entry for +$18.40. Yesterday's claims print helped in the most direct way available to this position: 203,000 initial claims against roughly 208,000 expected, with the insured unemployment rate unchanged at 1.2%. Getting the unemployment rate from 4.1% to 5.0% by December requires roughly 1.4 million additional unemployed in four months, and a labour market generating 203k weekly claims is not producing that. At 10:00 ET today the BLS publishes the preliminary annual benchmark revision to establishment payrolls, at the same hour Warsh gives his Jackson Hole keynote. We are writing six hours early, so this is a preview. The precedent worth holding: last year's benchmark took 2025 payrolls down by 911,000, and Cleveland Fed research notes the resulting -0.54% move sat only marginally outside the BLS's normal -0.5% to +0.5% range. So a large negative print is both plausible and less dramatic in context than the headline will read. The distinction that IS this position, restated on the day it matters: a benchmark revision restates establishment-survey payroll LEVELS. This contract settles on the household-survey unemployment RATE. Different surveys. The risk we accept, and have accepted since May, is that the market may not observe the distinction and a frightening headline moves the price without moving the probability.
▵ Bull case
- The contract settles on the household-survey unemployment rate; the benchmark revision restates establishment-survey payroll levels. Different surveys.
- Claims 203k versus ~208k expected, insured unemployment rate steady at 1.2% - no deterioration in the weekly data.
- Reaching 5.0% by December needs roughly 1.4 million additional unemployed in four months.
- GDPNow at 4.6% for Q3 describes an accelerating economy, which makes a 0.9 point rise in unemployment harder to construct, not easier.
▿ Bear case
- A 900k-plus downward benchmark revision produces alarming headlines, and prices move on headlines whether or not the underlying probability moves.
- Warsh speaking at the same hour doubles the chance of a disorderly reaction in a leg that only does ~$124K of volume.
- We publish before both events, so we cannot react today even if the revision genuinely changes the labour-market picture.
- At 11.2c the remaining edge to our FV of 8 is 3.2 points on a position already carrying a large unrealised gain - most of the value has been harvested.
4
Fed Rate Hike in 2026 - raising fair value into an event we will not see — YES
↑ BUY YES-2.5pp
Market price
57.5%
Fair value
55%
Gap: -2.5pp
We cut fair value to 53 on Tuesday largely because Brent had fallen roughly 9% from Friday, and we restored it to 54 on Wednesday only partially because hot PCE and cheap crude genuinely disagreed. Today the disagreement narrowed from the crude side: Brent settled $88.52 on Thursday, up 0.77% from Wednesday's $87.84, and trades $88.10 live this morning. The collapse stopped collapsing. Oil is roughly back where it was before the move that drove the cut, which removes an autumn disinflation impulse rather than adding one. Combine that with claims at 203k and a Cleveland August core nowcast holding at 3.40%, and the case for a 2026 hike is modestly better than it was 24 hours ago. FV 54 to 55. We should be explicit about what we are doing: raising a fair value six hours before the single largest scheduled event for this contract, which we will not be able to watch. Warsh delivers his first Jackson Hole keynote at 10:00 ET, and the risk we have flagged since Monday is unchanged. He began his term with an explicit bias against forward guidance, so he is unlikely either to promise a hike or to rule one out. The specific bad outcome for us is not dovishness. It is a hawkish chair who prefers accelerated balance-sheet runoff to rate increases, which is bearish for a contract that pays only on the rate.
▵ Bull case
- Brent settled UP 0.77% on Thursday at $88.52, taking back part of the crude-driven disinflation that justified Tuesday's cut.
- Claims 203k against ~208k expected; a tight labour market keeps a hike on the table.
- Cleveland August core PCE nowcast holding 3.40% YoY, with July core at 3.3% and headline 3.7%, both above published consensus.
- Polymarket's own September leg prices a 30.5c chance of a 25bp increase, roughly in line with the stale 31.6 FedWatch figure we have been carrying - the two independent reads agree.
▿ Bear case
- Warsh speaks in six hours and we publish before it. A preference for balance-sheet runoff over rate increases is hawkish in tone and bearish for this contract.
- Real consumer spending was flat in July; an economy losing momentum does not get hiked into.
- At 57.5c against our 55 FV we are marked 2.5 points rich on our own number, and have been marked rich here for several sessions.
- FedWatch has been unverifiable at source for an eighth consecutive session, so one of our two cross-checks on Fed pricing remains stale and dated.
5
Fed Rate End 2026 = 4.0% - the leg that is not allowed to have its own opinion — YES
↑ BUY YES+0.35pp
Market price
30.65%
Fair value
31%
Gap: +0.35pp
FV goes 30 to 31, derived mechanically from the pos-010 restoration 54 to 55, exactly as the previous two moves in this leg tracked pos-010 in both directions. We keep saying this out loud because it is the discipline that matters most in a book this correlated: the 4.0% bucket IS the one-hike outcome, so marking it independently would let us tell two different stories about a single event. On the live price we are about a third of a point cheap, which is noise, not edge. The useful thing today is that the whole ladder is verified fresh for the first time in three sessions: 3.5% at 7.55, 3.75% at 33.7, 4.0% at 30.65, 4.25% at 13.45 and 4.5%-or-higher at 4.4. That distribution has the market putting roughly 48 points on some tightening by year-end, which is coherent with the 57.5c on the standalone hike contract, and coherence between two related legs is worth checking even when it does not generate a trade. We are not trading it. A $25 ticket in a leg turning over a few hundred dollars a day costs more in spread to exit than a 0.35 point gap is worth.
▵ Bull case
- Mechanically consistent with pos-010 at FV 55, which is where the underlying work is done.
- Full ladder verified fresh: 3.5% 7.55 / 3.75% 33.7 / 4.0% 30.65 / 4.25% 13.45 / >=4.5% 4.4, coherent with the 57.5c standalone hike contract.
- The bucket is the direct expression of the single-hike scenario our main position is built on.
- At $25 the position is small enough that being marked a third of a point either way is immaterial to the book.
▿ Bear case
- The FV move is inherited, not independently derived, so any error in pos-010 propagates here undetected.
- A 0.35 point gap is not edge and we should not present it as one.
- Exit spread on a leg this thin exceeds the gap, so the position is effectively locked until resolution.
- If Warsh pushes the committee toward balance-sheet tools, both this leg and pos-010 lose together - the correlation runs in the losing direction too.