Global Rates
Fed September hike
51.5c (was 30.5c)
Repriced 22 points in three hours on Warsh's keynote, faded to 42.5c on Saturday, rebuilt over the weekend. Bid 51 / ask 52 on $13.9M. The single largest move on this board since we started publishing it.
CME FedWatch
about 57% - FRESH
First verifiable reading in nine sessions. The stale 31.6 we have been carrying and dating since Aug 20 is retired. Polymarket at 51.5c is 5.5 points below it, which is a real gap and still below our 10-point bar.
US 10Y / 2Y
4.73% / 4.34% (Aug 28 SETTLE)
Primary source: Treasury's daily par yield curve. The 2-year up 14bp against the 10-year's 6 is a hawkish flattening - the bond market pricing policy, not term premium. Live around 4.72% this morning.
European long end
Bund highest since 2011
CARRIED from Friday and quoted qualitatively rather than to four decimals, because the yield feed has not been re-observed. French yields the highest since 2009; UK 30-year near May peaks that were the highest since 1998.
BoJ September hike
88.0c
July unemployment 2.4%, a one-year low, and Tokyo August inflation at a five-month high. The data says hike and the leg agrees. Watchlist only - refused on correlation, not on price.
pos-017 fair value
61 (FROZEN), model 76.8
Barrier 7.0bp with 84 sessions left. Fourth consecutive failure of the 5.00% check, this time by 20.2 points. The freeze costs 15.8 points and is kept, with a pre-committed replacement published today.
The global long end has been selling off together for a fortnight and on Friday the United States finally led it. Kevin Warsh used his first Jackson Hole keynote to say that inflation is still too high and that recent cooler readings do not tell him underlying trends have meaningfully improved, and the market read that as a hiking chair. The two-year Treasury settled 4.34%, up 14 basis points; the ten-year settled 4.73%, up 6. Polymarket's September hike contract went from 30.5c at 14:00 UTC to 52.5c at 17:00 UTC and trades 51.5c now. CME FedWatch reads about 57%, which is its first fresh number in nine sessions - the stale 31.6 we have been carrying and labelling since Aug 20 is retired today. Everything on this board moved and we were not there for any of it, because this letter publishes at 04:06 ET and the speech was at 10:00. The three central-bank legs now read: ECB September 25bp increase 98.35c, BoJ September 25bp increase 88.0c, Fed September 25bp increase 51.5c against 47.5c for no change. The Fed leg is the only one of the three that is not already a foregone conclusion, and it is the only one we would consider - and we are refusing it, at a 5.5-point gap to FedWatch, because it is below our 10-point entry bar and because it would be a sixth expression of hawkish-Fed risk in a five-position book. The European and Japanese context is unchanged in direction and we are quoting it qualitatively today rather than to four decimals, because our yield feed for Bund, JGB and gilt has not been re-observed since Friday and we would rather label a carry than dress one up. The Bund is at its highest since 2011, French yields the highest since 2009, and UK 30-year borrowing costs are near the May peaks that were themselves the highest since 1998. In Japan the ten-year JGB is around 2.8%, July unemployment fell to 2.4% - a one-year low - and Tokyo inflation accelerated to a five-month high in August. That is a BoJ that hikes, which is why the September leg sits at 88c and why we will not buy it. For pos-017 all of this is the environment the position was designed for and none of it changes the mark. The barrier is 7.0 basis points, the model says 76.8, the fair value stays FROZEN at 61, and the 5.00% consistency check failed for a fourth consecutive session at 20.2 points - its widest yet. The freeze now understates the position by 15.8 points and we are keeping it, with a replacement rule pre-committed below before we know tomorrow's prices.
Today's Market Moves
Fed September 25bp increase
30.5%→51.5%+21.0pp
Flat at 30.5c until 14:00 UTC, 47.5c an hour later, 52.5c by 17:00. Bid 51 / ask 52 on $13.9M of volume against $15.9M on the no-change leg at 47.5c. A genuine coin flip on the most important scheduled macro event of the autumn.
ECB September 25bp increase
97.6%→98.35%+0.75pp
Effectively decided. Bid 98.2 / ask 98.5 on $104K. There is no trade here and the information value is that a September ECB hike is no longer a question, which is part of why the whole European curve keeps selling off.
BoJ September 25bp increase
87.5%→88.0%+0.5pp
Bid 87 / ask 89. Japanese July unemployment at a one-year low of 2.4% and Tokyo August inflation at a five-month high both point the same way. We mark it about fair and refuse it on correlation, which is the only honest reason available.
10Y Touches 4.8% Before 2027 (pos-017)
64.5%→67.5%+3.0pp
Bid 65 / ask 70. The written exit gate is a 66c BID and the bid is 65 - one cent. Correction to Friday's letter: we said the gamma feed returns last and mid but not the bid. It returns bestBid and bestAsk, and always did. We were wrong about our own data source, on the one field the gate depends on.
10Y Touches 5.00% Before 2027 (check leg)
17.5%→21.5%+4.0pp
The diagnostic. Our model on the same inputs says 41.7 against a live 21.5c with a 20 bid. Fourth consecutive failure and the widest yet at 20.2 points; the sequence is 14.1, 17.0, 17.6, 20.2. One direction, growing.
Fed Rate Hike in 2026 (pos-010)
57.5%→67.5%+10.0pp
Our position, and the reason the global board matters to this book. 56.5c at 14:00 UTC to 68.5c at 16:00. Fair value 55 -> 67. The trim gate above 72c missed firing by 3.5 cents, which we are recording as luck rather than process.
Screening Table
| # | Market | Expiry | Market Price | Fair Value | Gap (pp) | Direction | Volume | Confidence |
|---|---|---|---|---|---|---|---|---|
| 1 | 10Y Touches 4.8% Before 2027 (pos-017) | Dec 31 | 67.5% | 61% | -6.5pp | HOLD $25 YES - model says 76.8, our rule freezes it at 61, the check failed a 4th time; exit gate needs a 66c bid and the bid is 65 | $$60K on the leg | 3/10 |
| 2 | Fed September 25bp increase | Sep 16 | 51.5% | 57% | +5.5pp | NO POSITION - a real 5.5pp gap to a FRESH FedWatch near 57, but below our 10pp bar and a 6th correlated hawkish-Fed bet | $$13.9M on the leg | 3/10 |
| 3 | BoJ September 25bp increase | Sep 18 | 88.0% | 88% | 0.0pp | NO POSITION - marked fair on a 2.4% unemployment rate and five-month-high Tokyo inflation; refused on correlation alone | $$127K on the leg | 3/10 |
| 4 | ECB September 25bp increase | Sep 10 | 98.35% | 98% | -0.35pp | NO POSITION - decided; no edge worth a ticket and none claimed | $$104K on the leg | 4/10 |
| 5 | 10Y Touches 5.00% Before 2027 | Dec 31 | 21.5% | 21.5% | 0.0pp | NO POSITION - our model says 41.7, a nominal 20.2pp gap, REFUSED for a 4th session; we publish our mark AT the market because that gap IS our error term | $$83K on the leg | 1/10 |
Top 5 Opportunities
1
Fed September - the board repriced 22 points in three hours and we were already published — NO POSITION
↑ BUY YES+5.5pp
Market price
51.5%
Fair value
57%
Gap: +5.5pp
The hourly record is worth printing because it is unusually clean. The September hike leg traded 30.5c from midnight UTC through 14:00. At 15:00 it was 47.5c. At 16:00, 50.5c. At 17:00, 52.5c. It then faded to 42.5c through Saturday and rebuilt across the weekend to 51.5c, bid 51 / ask 52, on $13.9M. The no-change leg is 47.5c on $15.9M. This is now a coin flip on the most consequential scheduled macro event of the autumn. Two months ago this letter was carrying a FedWatch reading of 31.6 that we could not verify at source, dating it, and calling it stale in every issue rather than quietly refreshing it. That ends today: FedWatch is readable again at roughly 57% for the September meeting, up about 20 points on Warsh's remarks and up from around 40% a week ago. So the futures-implied probability is 57 and the prediction market is 51.5, a gap of 5.5 points in the direction that would pay us. We are not taking it. The gap is below our 10-point entry bar, and more importantly this book already holds four separate expressions of the same idea - a hike in 2026, a 4.0% end-2026 rate, zero cuts, and a 4.80% ten-year touch. Adding the September leg would make five out of six positions the same bet with different labels. The correlation cap is only a real constraint on days when the correlated trade looks good, and today it looks very good.
▵ Bull case
- FedWatch near 57 against a 51.5c market is a 5.5-point gap on a deeply liquid leg, and prediction markets have historically converged toward the futures strip into a meeting.
- The chair said it himself, in prepared remarks, at the Fed's own conference. That is the highest-quality possible input for a contract on the Fed's own decision.
- The two-year moved 14 basis points on the day, which is the rates market putting real money behind the same reading rather than a headline interpretation.
- ECB September is 98.35c and BoJ September 88c. A Fed that does not move while both peers tighten has a currency problem as well as an inflation problem.
▿ Bear case
- 5.5 points is below our published 10-point bar and we do not get to move the bar because we like the trade.
- It would be a sixth hawkish-rates position in a five-position book. That is not diversification, it is leverage on one idea.
- The August employment report lands Friday Sep 4, before the meeting, with consensus at +60,000 and the unemployment rate seen ticking to 4.2%. A soft print collapses this leg and everything correlated with it at the same time.
- The leg already round-tripped from 52.5c to 42.5c and back inside a weekend. That is not the volatility profile of a settled view.
2
10Y Treasury Touches 4.80% Before 2027 - seven basis points away, and we froze the mark again — YES
↑ BUY YES-6.5pp
Market price
67.5%
Fair value
61%
Gap: -6.5pp
Everything this position was built for happened at once. The ten-year settled 4.73% against a 4.80% touch level, so the barrier is 7.0 basis points with 84 bond-market sessions left in the year. The Bund is at its highest since 2011, French yields the highest since 2009, UK 30-year borrowing costs near the highest since 1998, and the chair of the Federal Reserve has told the market the next move may be up. The arithmetic on our published model says 76.8. The fair value stays at 61. The standing rule from Aug 25 blocks any increase until the 5.00% consistency check reproduces, and today it failed for a fourth consecutive session and by the widest margin yet: the same model, the same inputs, a 27.0 basis point barrier, 41.7 after the haircut, against a market trading 21.5c with a bid at 20. The failure sequence is 14.1, 17.0, 17.6, 20.2 points. Consistently one direction, and growing. A model that is twenty points rich on the far leg has not earned the right to be believed on the near leg. The cost of that discipline is now 15.8 points, up from 5.8 on Friday, and it produces an awkward reading: at 61 our own published fair value says this position is 6.5 points RICH at the 67.5 market, so on our own numbers it is a sell - while the written exit gate says sell on a 66c bid and the bid is 65. We are doing neither thing. We will not sell early because our fair value says so, and we will not move the gate because it is close. PRE-COMMITMENT, WRITTEN BEFORE WE KNOW TOMORROW'S PRICES: a fifth consecutive failure of the 5.00% check retires the barrier model from the published fair value - it stays as a diagnostic - and re-anchors pos-017 to the market mid less a two-point liquidity discount, effective at the next run whether that helps or hurts.
▵ Bull case
- Seven basis points, four months, and a global long end that has moved 6 to 14 basis points in a single session twice in the past week.
- The 4.6% leg on the same ladder has already resolved at 1.00. The touch distribution this year has been running ahead of what a fixed-vol model would have projected in January.
- Europe is doing the work independently of the Fed. A Bund at 2011 highs and gilts near 1998 highs means US 4.80% no longer needs a purely domestic story.
- ECB September is 98.35c and BoJ September 88c. Simultaneous tightening from all three major central banks is the regime in which long yields grind higher rather than mean-revert.
▿ Bear case
- Four consecutive failures of our own consistency check is the strongest available evidence that the model underneath this position does not work, and the model was the only source of claimed edge.
- Bid 65 / ask 70 on $60K. A five-point spread on a thin leg is a bad place to be confident, and it is why the exit gate is written on a bid rather than a mid.
- Our own frozen fair value says this is 6.5 points rich today. If we believed our published numbers we would already be out.
- The buyback window expires Nov 4 while the barrier runs to Dec 31, so the final eight weeks lose the supply-side support that formed part of the original thesis.
3
BoJ September - the data says hike, the leg says hike, and we still will not buy it — NO POSITION
↑ BUY YES0.0pp
Market price
88.0%
Fair value
88%
Gap: 0.0pp
Japanese July unemployment fell to 2.4%, a one-year low, and Tokyo inflation accelerated in August to a five-month high. The ten-year JGB is around 2.8%. Those three facts together describe a central bank that has run out of reasons to wait, and the September 25bp increase leg agrees at 88.0c, bid 87 / ask 89. We mark it fair. That is the whole analysis, and it is deliberately short, because the interesting question here is not the price - it is why a position we think is fairly priced still gets refused. This book holds four expressions of hawkish developed-market rates risk and a fifth would correlate with all of them: a BoJ hike lifts JGB yields, JGB yields drag the global long end, the global long end is exactly the channel through which pos-017 wins. On a good day that is four winners at once. On a bad day it is four losers at once, in a book with $332 at risk. The reason we write this leg up every day without buying it is that the correlation cap should be visible when it costs something. Today it costs something.
▵ Bull case
- July unemployment at 2.4% is a one-year low and Tokyo August inflation is at a five-month high. Both series point the same direction into a September meeting.
- The BoJ has been reducing its JGB holdings, and finding new buyers for the long end is the structural problem behind the whole Japanese curve. Normalising the policy rate is part of that answer.
- The ECB is at 98.35c for September. A BoJ that stands still while the ECB tightens reopens the yen weakness problem that pushed it to move in the first place.
- At 88c with a 2-point spread the leg is liquid enough to trade in size we would actually use.
▿ Bear case
- We mark it at 88 against an 88 market. There is no edge, so the correlation argument is not even the binding constraint - it is simply not a trade.
- It would be a fifth correlated position and the book is already saturated with the same idea expressed four ways.
- Japanese data can turn on one soft Tokyo CPI print, and the leg has almost no downside cushion at 88c.
- We have no independent Japanese model. Marking it fair is really marking it to the market, and we would rather say that than imply analysis we have not done.
4
ECB September - decided, and the decision is itself the information — NO POSITION
↑ BUY YES-0.35pp
Market price
98.35%
Fair value
98%
Gap: -0.35pp
98.35c, bid 98.2 / ask 98.5. There is no trade here and we are not going to invent one. What the level tells you is that a September ECB hike has stopped being a question, and that matters for the rest of this board more than the leg itself does. The European long end is where the global bond selloff is currently most extreme - the Bund at its highest since 2011, French yields the highest since 2009, UK 30-year borrowing costs near the May peaks that were the highest since 1998. A central bank universally expected to tighten into that is not going to be the thing that stops it. That is a supporting argument for pos-017 and we count it as one, while noting that it is a supporting argument and not an independent one, since it is the same rates idea we already own four times. Housekeeping: our Bund, JGB and gilt levels have not been re-observed since Friday, so today we describe them qualitatively rather than quoting four decimal places we have not checked. A carried number that looks precise is worse than a carried number that admits it.
▵ Bull case
- At 98.35c the market has effectively removed September from the set of open questions, which raises the informational value of October, currently the more interesting contract.
- Euro-area flash CPI for August lands this week and is the last input before the Sep 10 decision.
- A tightening ECB keeps upward pressure on the Bund, which is the clearest transmission channel into the US long end and therefore into pos-017.
- The leg is well behaved and liquid enough to be a reliable read on European policy expectations, which is the use we make of it.
▿ Bear case
- 1.65 points of upside. There is no version of this that is worth a ticket, and pretending otherwise would be the kind of manufactured edge this letter exists to avoid.
- Correlated with everything else we own, like the rest of this board.
- We do not have an independent European inflation model, so our 98 is really the market's 98.35 rounded down. We should say so.
- A surprise soft euro-area flash CPI this week is the one thing that could make this leg interesting, and by then it would be interesting to everybody.
5
10Y Touches 5.00% - a 20.2pp model gap declined for a fourth time, and marked at the market instead — NO POSITION
↑ BUY YES0.0pp
Market price
21.5%
Fair value
21.5%
Gap: 0.0pp
Our barrier model, run on the same inputs that produce a 76.8 for the 4.80% leg we own, says this contract is worth 41.7. It trades 21.5c, bid 20 / ask 23. That is a nominal 20.2-point gap, the largest on this board and larger than anything we have taken all year. We are declining it for a fourth consecutive session, and the reason is that this gap is not an opportunity - it is our error term. The check exists precisely to catch this. If a model claims a twenty-point edge on a far barrier in a market with $83K of volume and real participants on both sides, the likeliest explanation is not that the market is asleep; it is that a fixed-volatility barrier model over-prices far touches by construction, because it assumes a constant daily sigma of 4.00 basis points all the way to Dec 31 and ignores that big moves cluster and then stop. The failure has now been recorded four times: 14.1 points, 17.0, 17.6, 20.2. Every one in the same direction and each larger than the last. That pattern is the entire justification for the pos-017 freeze, and it is why we have pre-committed today to retiring the model from the published fair value if it fails a fifth time. Publishing a refused twenty-point gap costs us nothing except the temptation to take it, and that is the point.
▵ Bull case
- The leg has risen from 17.5c to 21.5c since Friday, so the market is moving toward the model rather than away from it.
- If the ten-year touches 4.80% - and it is 7 basis points away - the conditional probability of 5.00% by December is materially higher than 21.5c implies.
- Global long-end pressure is real and simultaneous across the US, Germany, France, the UK and Japan, which is the regime in which far barriers get reached.
- $83K of volume is thin, and thin markets are sometimes genuinely mispriced rather than just risky.
▿ Bear case
- Our own consistency check says this specific number is where our model breaks. Buying the thing your diagnostic says is broken is not edge, it is a category error.
- Fixed-volatility barrier models over-price far touches structurally. The 4.00bp daily sigma assumption is doing all the work and it is an assumption, not an observation.
- It is a fifth correlated rates position in a book already carrying four.
- Bid 20 / ask 23 on $83K. Even if the edge were real, three points of spread on a thin leg eats a meaningful share of it.