Daily Macro US
pos-017 CLOSED
+$83.61 realised, sold at 71.69c
The 66c-bid take-profit fired at an 82c bid and we sold all 151.5152 shares. Volume-weighted fill 71.69c: the book paid 10 shares at 82c, 10 at 81c, 3.53 at 73c, 14.82 at 71c and 113.17 at 70c. Every share above the trigger. Proceeds $108.61 on a $25 stake.
Marking convention CHANGED
BID, not mid, from today
Friday's mid marked pos-017 at $132.58. The realisable value was $108.61. We carried $23.96 of profit that did not exist at our size. The new convention lowers our own reported open book by $5.36 - it is a change that only ever costs us.
10Y Treasury
4.75% (Aug 31 SETTLE)
Treasury's daily par yield curve, primary source. Up 2bp, highest since January 2025; 2-year flat at 4.34%, 30-year 5.25%. The barrier to a 4.80% touch is 5.0 basis points.
Global long end
JGB 3.00%, Bund 3.3%
The JGB 10-year touched 3% for the first time since 1996 and the Bund hit 3.3% for the first time since May 2011, with French yields at 2008 highs. This is the move that repriced pos-017 twenty points overnight.
5.00% consistency check
11.1pp - 5th straight failure
Model 44.6 against a 33.5c market. The gap NARROWED from 20.2 and still cleared the bar. The Aug 31 pre-commitment retires the barrier model to diagnostic-only, and we wrote it in advance precisely so the narrowing could not be used as an excuse.
CME FedWatch, September hike
66% (Aug 31)
Up from about 57 on Friday and the second consecutive fresh reading after nine stale sessions. Polymarket's September leg is 56.5c, so the futures market is 9.5 points ABOVE the prediction market - under our 10-point bar, and it would be a sixth correlated hawkish-Fed bet.
The take-profit gate we published every day for a fortnight finally fired, and the number it produced is not the number the screen showed. pos-017's rule was sell on a 66c BID. On Friday the bid was 65 and we wrote that we were not pre-empting it and not moving it. Overnight it went to 82. So we sold - all 151.5152 shares of it - and the volume-weighted fill was 71.69c, because the order book had ten shares at 82c, ten more at 81c, and then a hole. The realised gain is +$83.61 on a $25 stake, a 334% return and the second-largest winner this book has produced. Every share filled above the 66c trigger, which is what makes it a clean execution rather than a slippage story. But here is what it cost us to learn. We have been marking this position at the mid. Friday's mid of 87.5c said the position was worth $132.58. The realisable value at our size was $108.61. We were carrying $23.96 of paper profit that was never there, on a $25 ticket, for weeks - and the only reason we found out is that we finally tried to sell into a book instead of reading a price off it. FROM TODAY THIS BOOK MARKS EVERY POSITION AT THE BID. That change lowers our own published open P&L by $5.36 on the four remaining positions, which is the only sort of methodology change a reader never has to take on trust. A SECOND GATE FIRED AND THIS ONE SENT A BILL. pos-010's trim above 72c triggered at 75.5c at 17:00 UTC on Monday, in the middle of the US session, five hours after Friday's letter published. It is 71.5c now. Under the rule we adopted on Aug 31 - a gate that fires in our absence gets acted on at the next run and never quietly re-dated - we trimmed half the remainder this morning at the 70c bid. The gate fired at 75.5 and we executed at 70: 5.5 points, $5.00 on this half, surrendered to the gap between when the rule triggered and when we were next at the desk. Friday we called a 3.5-cent near-miss luck. This is not luck, it is a bill, and it is the second time in three sessions that publishing at 04:00 ET has cost this book real money. And there is a sharper edge on it: our fair value on pos-010 is 71, so the gate made us sell one point BELOW our own number. We honoured it anyway, because a gate you only obey when it agrees with your fair value is decoration - which is precisely what we said about pos-011's old 40c trigger on Saturday. THE PRE-COMMITMENT ALSO FIRED, ON SCHEDULE. On Aug 31, before we knew today's prices, we wrote that a fifth consecutive failure of the 5.00% consistency check would retire the barrier model to diagnostic-only. It failed a fifth time this morning at 11.1 points. It also NARROWED sharply, from 20.2, and that is exactly the argument we wrote the pre-commitment to stop ourselves from making. The model is retired. Behind all of this the world sold bonds together: the JGB 10-year touched 3.00% for the first time since 1996, the Bund reached 3.3% for the first time since May 2011, French yields hit their highest since November 2008, and the US 10-year settled 4.75%, its highest since January 2025, on renewed US-Iran strikes that put Brent above $91. Written before ISM manufacturing at 10:00 ET, and three days before the August employment report.
Today's Market Moves
10Y Touches 4.8% Before 2027 (pos-017)
67.5%→87.5%+20.0pp
THE POSITION WE JUST SOLD. Bid 82 / ask 93 on a mid of 87.5 - and the bid had ten shares behind it. Hourly path: 67.5c at 12:00 UTC Monday, 76.5c at 14:00, 78c at 21:00, 85.5c at 02:00 this morning, peaking 88.5c at 05:00. The take-profit gate needs a 66c bid; we cannot reconstruct the historical bid series, only the mid, so we say plainly that the mid crossed 66c during Monday's US session and the bid almost certainly followed hours before we saw it.
Fed Rate Hike in 2026 (pos-010)
67.5%→71.5%+4.0pp
Our position, half of the remainder now trimmed. Hourly: 67.5c at 13:00 UTC Monday, 74c at 14:00, 72.5c at 15:00, 75.5c at 17:00, 72c by 19:00 - three hours above the 72c trim gate while this letter was five hours old. Executed today at the 70c CLOB bid, where the book shows 5,681 shares. Gamma reports bestBid 71; we used the level with confirmed depth.
US Unemployment 2026 >= 5.0% (pos-004)
10.65%→7.1%-3.55pp
Bid 6.1 / ask 8.1. The early-exit review armed below 8c FIRED, in daylight this time: 10.65c through 13:00 UTC, 8.75c at 17:00, 7.60c at 18:00, 7.05c from 19:00 onward. Reviewed below; the answer is hold. Note that at 7.1c the market is now BELOW our fair value of 8, so the NO we hold is marked rich rather than cheap for the first time since entry.
10Y Touches 5.00% Before 2027 (check leg)
21.5%→33.5%+12.0pp
The diagnostic, never a position. Bid 28 / ask 39. Our barrier model on the same inputs says 44.6 - a fifth consecutive failure at 11.1 points, and the first NARROWING in the sequence (14.1, 17.0, 17.6, 20.2, 11.1). The pre-commitment written Friday retires the model anyway. A rule you rewrite the moment it starts to hurt was never a rule.
Zero Fed Rate Cuts in 2026 (pos-013)
87.95%→88.95%+1.0pp
Bid 88.9 / ask 89.0 on $7.76M, one tick wide. The 90c trim did NOT fire: the hourly high since Monday midnight UTC is 88.95c, printed at 08:00 this morning, and it is now 1.05c away - closer than at any point in this position's life. Fair value 90 -> 91 on FedWatch at 66% for a September hike.
Fed September 25bp increase
51.5%→56.5%+5.0pp
Bid 56 / ask 57 on $15.1M; 'no change' is 42.5c. FedWatch reads 66%, so the gap is 9.5 points in the prediction market's favour - below our 10-point bar for the second session running, and it would be a sixth expression of the same hawkish-Fed view in what is now a four-position book. Refused on both counts.
Screening Table
| # | Market | Expiry | Market Price | Fair Value | Gap (pp) | Direction | Volume | Confidence |
|---|---|---|---|---|---|---|---|---|
| 1 | 10Y Touches 4.8% Before 2027 (pos-017) | Dec 31 | 87.5% | retired% | 0pp | SOLD $25 YES at a volume-weighted 71.69c - the 66c-bid gate fired at an 82c bid; realised +$83.61 | $$60.8K on the leg | 5/10 |
| 2 | Fed Rate Hike in 2026 (pos-010) | Dec 31 | 71.5% | 71% | -0.5pp | TRIMMED half the remainder at 70c on the 72c gate - one point BELOW our own fair value, and honoured anyway | $$8.32M | 4/10 |
| 3 | Zero Fed Rate Cuts in 2026 | Dec 31 | 88.95% | 91% | +2.05pp | HOLD $82 YES - FV 90 -> 91 on FedWatch at 66% for September; the 90c trim did not fire and is 1.05c away | $$7.76M | 4/10 |
| 4 | Fed Rate End 2026 = 4.0% | Dec 31 | 38.75% | 44% | +5.25pp | HOLD $25 YES - mechanical FV 44, but the ladder and the hike contract now disagree by 9.3 points, so treat the 5.25 with that much doubt | $$1.37M on the leg | 2/10 |
| 5 | US Unemployment 2026 >= 5.0% | Dec 31 | 7.1% | 8% | -0.9pp | HOLD $100 NO - the sub-8c review fired and concluded hold; $10.80 of upside left, all of the edge inside the spread | $$124.6K on the leg | 5/10 |
Market vs Fundamentals
Market Price (red) vs Estimated Fair Value (green) — %
Top 5 Opportunities
1
10Y Touches 4.8% Before 2027 - the gate fired, and the bid had ten shares behind it — YES
↑ BUY YES0pp
Market price
87.5%
Fair value
0%
Gap: 0pp
We published this gate every day for two weeks: sell on a 66c BID. On Friday the bid was 65 and we wrote that we were not pre-empting it and not moving it. This morning it is 82, so we sold, and the execution is published as a ladder because the ladder is the point. 151.5152 shares from a $25 stake at 16.5c. The book paid 10 shares at 82c, 10 at 81c, 3.53 at 73c, 14.82 at 71c, and the last 113.17 at 70c. Proceeds $108.61, volume-weighted 71.69c, realised +$83.61 - a 334% return and the second-largest winner this book has produced. Every single share filled above the 66c trigger, which is the only thing that makes this a clean execution rather than an excuse. Now the uncomfortable half. We have marked this position at the mid since we opened it. Friday's 87.5c mid valued it at $132.58. The realisable value at our size was $108.61. That is $23.96 of paper profit that was never available to us, carried for weeks on a $25 ticket, and we discovered it only because we finally sold into a book instead of reading a number off a screen. Every prediction-market portfolio that marks at the mid has this problem; ours had it too, and we are the ones who have to say so.
▵ Bull case
- The thesis was correct. The 10-year settled 4.75% on Aug 31, its highest since January 2025, leaving a 5.0 basis point barrier to the 4.80% touch. We bought this at 16.5c in July on the view that the long end was not done selling off.
- The move that got us here was global rather than American: the JGB 10-year touched 3.00% for the first time since 1996 and the Bund reached 3.3% for the first time since May 2011. A synchronised long-end selloff is a much better reason for a US touch than anything domestic.
- Every share filled above the written trigger. A gate whose execution range sits entirely above its trigger price is a gate that worked, whatever the average.
- Closing this removes the book's single largest correlated bet on the hawkish-rates theme, which is the diversification we have been talking about for three weeks without being able to act on.
▿ Bear case
- We sold at a volume-weighted 71.69c into a market whose mid was 87.5c. If the 10-year touches 4.80% - and it is five basis points away - we will have surrendered roughly $43 of the $133 this position would have paid.
- The mid-marking error is ours and it flattered this book for weeks. We do not get credit for finding it; we get the correction and the new convention.
- We were late. The bid crossed 66c during Monday's US session, hours before we saw it, and the fill we got is the fill available at 04:00 ET the next morning rather than the one available when the gate fired.
- The barrier model that justified this position for two months is being retired today for failing its own consistency check five times. It happened to be right here. That is not the same as it having been reliable.
2
Fed Rate Hike in 2026 - the gate made us sell below our own fair value and we did it — YES
↑ BUY YES-0.5pp
Market price
71.5%
Fair value
71%
Gap: -0.5pp
The trim gate above 72c has been published since June. On Monday the leg traded 74c at 14:00 UTC, 72.5c at 15:00 and 75.5c at 17:00 - three hours above the trigger, in the middle of the US session, five hours after Friday's letter had already gone out. It is 71.5c now. The rule we adopted on Aug 31 says a gate that fires in our absence is acted on at the next run and never quietly re-dated, so we trimmed half the remainder this morning: 90.9091 shares at the 70c CLOB bid, where the book shows 5,681 shares of depth. Proceeds $63.64 against $50 of stake, realised +$13.64. The gate fired at 75.5 and we filled at 70. That 5.5 points is $5.00 on this half and it is a bill, not noise - the second time in three sessions that a 04:00 ET publish time has cost this book real money, and unlike Friday's 3.5-cent near-miss it is not something we get to call luck. There is a sharper edge still. Our fair value here is 71 and we sold at 70. The gate made us sell below our own number. We honoured it anyway, because a gate that binds only when it agrees with your fair value is decoration - which is exactly what we said about pos-011's old 40c trigger two days ago, and consistency is not optional when the inconsistency would be profitable.
▵ Bull case
- CME FedWatch has a September hike at 66%, up from about 57 on Friday, and it is the second consecutive fresh reading after nine stale sessions. The futures market is more convinced than the prediction market.
- The global long end broke together: JGB 3.00% for the first time since 1996, Bund 3.3% since May 2011, French yields at 2008 highs. Every central bank facing that curve has a harder disinflation argument.
- Renewed US-Iran strikes - Larak Island, Iranian attacks on the UAE and Jordan, Trump threatening Kharg Island, a supertanker burning in Hormuz after hitting two mines - put Brent above $91. The crude disinflation we cut positions for on Aug 26 has fully reversed.
- Cleveland's August core PCE nowcast re-verified at source this morning, unchanged at 3.40% year-over-year with headline at 3.79%, stamp 08/31. That is not a committee's cutting environment.
▿ Bear case
- We sold at 70 against a fair value of 71. If this is a good position, we have just made it smaller for a point.
- Inflation-adjusted consumer spending was FLAT in July after +0.4% in June, and July payrolls FELL 23,000. The mechanism by which a hike becomes impossible is still intact and nothing this week touched it.
- Friday's August employment report - consensus +60,000 with the unemployment rate expected at 4.2% from 4.1% - can undo most of this in one print, and it lands three days from now.
- At 71.5c against a fair value of 71 there is nothing left to collect on the remainder either. We are holding $50 of a position that is priced correctly, which is a decision we should keep re-justifying.
3
US Unemployment 2026 >= 5.0% - the review fired, and we refused to move the trigger this time — NO
↑ BUY YES-0.9pp
Market price
7.1%
Fair value
8%
Gap: -0.9pp
The early-exit review armed below 8c fired at 17:00 UTC Monday and this one we could actually watch happen: 10.65c through 13:00, 8.75c at 17:00, 7.60c at 18:00, 7.05c from 19:00 and flat all night. The published wording has always been that this triggers a review, not an automatic sale, so here is the review. $100 of NO at 75c is 133.3333 shares. At the 91.9c NO bid a sale realises $122.53 today; a correct resolution in December pays $133.33. We would be banking $22.53 now to surrender $10.80 of pull-to-par over four months. Our fair value on the YES is 8, so the fair NO is 92.0 against a 91.9 bid: the entire edge is one tenth of a point inside a two-point spread. That is the same shape as pos-011's review on Saturday and it gets the same answer, because handling two identical situations differently is how a process quietly stops being one. What we are NOT doing is raising the review line. On Aug 31 we moved pos-011's trigger up on the day it fired and told readers to be suspicious of that pattern; doing it again three sessions later would prove them right. The 8c line stays, the review re-runs every session the leg is under it, and we add an action-trigger beneath it instead: YES below 5c closes this position automatically, no review and no discretion.
▵ Bull case
- Getting from 4.1% to 5.0% by December still requires roughly 1.4 million additional unemployed Americans in four months. Nothing in the data flow makes that a live scenario.
- The BLS benchmark payroll revision on Friday landed at -79,000, or -0.1%, against last year's -911,000. The frightening headline this position was exposed to did not arrive.
- A Fed that FedWatch puts at 66% to hike in September is not a Fed looking at a collapsing labour market.
- At the 91.9c bid we would be selling at exactly our own fair value of 92.0. There is no efficiency gain available, only a spread to pay.
▿ Bear case
- July payrolls FELL 23,000. That is the genuine bear datapoint and it gets full weight, not a footnote.
- Friday's consensus has the unemployment rate ticking to 4.2% from 4.1%. Two more prints like that and the market re-rates this leg, and our mark goes with it.
- We are risking $122.53 of realisable value to collect $10.80. Stated as a ratio that is eleven to one against us, and no amount of confidence in the base rate makes that a comfortable sentence to write.
- At 7.1c the market is now cheaper than our fair value of 8, which means the NO we hold is marked rich. We are holding a position our own model says is slightly expensive.
4
Zero Fed Rate Cuts in 2026 - the trim is 1.05 cents away and we are not moving it either — YES
↑ BUY YES+2.05pp
Market price
88.95%
Fair value
91%
Gap: +2.05pp
The largest stake in the book gained a point and the case for it improved more than the price did. Fair value goes 90 to 91 on news rather than tape. CME FedWatch has a 25 basis point September HIKE at 66%: if the committee is seriously debating tightening in a fortnight, a cut in the four months after that stops being a judgement call and becomes close to arithmetic. Cleveland's August core PCE nowcast was re-verified at source this morning and is unchanged at 3.40% year-over-year with headline at 3.79%, stamp 08/31 - fresh, not carried. Brent is above $91 on renewed US-Iran strikes, which means the crude disinflation that made us cut this position on Aug 26 has now completely reversed. And the global long end broke to multi-decade highs together. On Friday we noted that this leg went the wrong way on its best news and said that if it happened twice the interpretation would change. It did not happen twice: the position went up a full point on a day when everything confirmed the thesis, so Friday's 0.75 reads in retrospect as the weekend drift we called it. The 90c trim did not fire - the hourly high since Monday midnight UTC is 88.95c, printed at 08:00 this morning - and it now sits 1.05 cents away, closer than at any point in this position's life. It stays exactly where it is.
▵ Bull case
- FedWatch at 66% for a September hike. The distance between 'might hike next meeting' and 'might cut before December' is not a distance this committee can travel in four months.
- Cleveland's August core PCE nowcast re-verified UNCHANGED at 3.40% year-over-year, headline 3.79%, stamp 08/31 - a fresh reading, not a carry.
- Brent above $91 on renewed US-Iran strikes. The Aug 26 crude collapse that gave the doves an argument has fully reversed within a week.
- JGB 3.00% for the first time since 1996 and Bund 3.3% for the first time since May 2011. Global term premium is repricing upward, which is the opposite of a cutting environment.
▿ Bear case
- Real consumer spending was FLAT in July after +0.4% in June. This is the mechanism by which the committee ends up cutting anyway, and it has not gone away.
- July payrolls FELL 23,000 and Friday's consensus has unemployment at 4.2%. One genuinely bad August print puts a 2026 cut back on the table, and it is three days out.
- At 88.95c the remaining upside is 11.05 cents on the book's largest stake, against a tail that a labour-market break would make very expensive very quickly.
- GDPNow is CARRIED at 4.6% from Aug 26 for a fourth session. The Atlanta Fed says it refreshes today, after ISM manufacturing, so the strongest growth number in our bull case is also the stalest.
5
Fed Rate End 2026 = 4.0% - our fair value is derived from a contract that now disagrees with itself by nine points — YES
↑ BUY YES+5.25pp
Market price
38.75%
Fair value
44%
Gap: +5.25pp
The one-hike bucket gave back two thirds of a point while the standalone hike contract gained four, and that divergence is the whole entry. The ladder prices 3.5% at 8.30, 3.75% at 25.15, 4.0% at 38.75, 4.25% at 18.00 and 4.5%-or-more at 5.45, so the hike mass sums to 62.20. The standalone Fed-hike-in-2026 contract trades 71.5. Two markets on one event, disagreeing by 9.3 points - up from 4.35 on Friday, so the inconsistency we flagged against ourselves has more than doubled in a session. Our fair value here is derived FROM the contract that disagrees, mechanically: 71 x 38.75/62.20 = 44.2, rounded to 44. So the 5.25-point gap on the screen carries 9.3 points of internal contradiction behind it, and we publish both numbers rather than the flattering one. Under today's new bid-marking convention this $25 position is worth $25.88 - eighty-eight cents of gain, against $3.49 if we had kept marking at the mid. A two-and-a-half-dollar difference on a $25 ticket is the clearest possible illustration of why the convention had to change, and it is not a coincidence that the position with the widest spread in the book is the one the old method flattered most.
▵ Bull case
- A mechanical derivation from a deep, one-tick-wide contract is a better estimator than a judgement call, and it says 44 against a 38.75 market.
- If the Fed hikes once in September and stops, this is the bucket that pays. FedWatch at 66% for September makes exactly that path the modal one.
- The ladder shifted the right way on the week even after Monday's drift: 4.25% went 13.45 to 18.00 and 4.5%-or-more 4.5 to 5.45 since Friday's letter, which is the distribution moving toward us.
- $25 of stake at 34c on a contract that pays $73.53 if correct is the kind of asymmetry that survives a wide spread.
▿ Bear case
- The 9.3-point disagreement between the ladder and the hike contract is 9.3 points of doubt attached directly to our fair value, because our fair value is computed from one of them.
- The spread is 7.1 points wide - bid 35.2, ask 42.3 - on a $25 ticket. The entire claimed 5.25-point edge fits inside it with room to spare.
- Marked at the bid this position has made eighty-eight cents in three months. That is not an edge, it is a rounding error with a story attached.
- Two hikes rather than one, or none at all, and this bucket pays nothing. It is the narrowest possible expression of a view we already hold twice elsewhere.