Pinned Tweet
Call: the four-session collapse in credit stress is the unwinding of the 6 November volatility spike, not an improvement in credit.
Credit & Stress is +0.308. On 6 November it was +0.960. That is 0.652 across four sessions, and across the 48 daily snapshots this desk holds no four-session stretch has produced a larger decline. The prior worst was 0.488.
Now the split, because the split is the whole argument. 105% of that fall came from the two insurance prices. VIX supplied 0.610, MOVE 0.074. The two default prices went the wrong way: high-yield OAS plus 0.022, investment grade plus 0.009.
VIX printed 17.49 from 28.18, a fall of 10.69 points. The S&P is still 2.18% below its 5 November close, so the equity market has not recovered what it lost; only the price of insuring against a repeat has.
⚠️ The spread levels did not confirm relief. High-yield OAS is 3.34% (as of 9 November), 14 basis points inside its 1yr high but in the 96.6th percentile of the year. Investment grade is 0.91%, 3 basis points off its own 1yr high.
And the 10yr rose 10.4 basis points over the same window, to 4.998%. A treasurer does not sign the spread. They sign the spread plus the Treasury. All-in high-yield cost is 8.338% against 8.294%, plus 4.4 basis points. Investment grade is plus 8.4.
So the gauge says the system got calmer and the invoice says borrowing got dearer. Both are true. That is what an unwind of a vol spike looks like when nothing was ever wrong with credit and nothing has been fixed either.
What kills the read: high-yield OAS tightening through 3.14% with investment grade below 0.86%, which puts the complex back where it was before the widening. Or the VIX above 24 with high yield still above 3.30%, which makes this a pause inside a volatility regime rather than an exit from one.
Composite is +0.245, NEUTRAL for a second session, on Scarcity at 1.115. macrotape.io/posts/2026-11-1…
🏭 Everyone is arguing about whether AI is a bubble. Almost nobody is counting what has actually been installed.
Nvidia's data centre revenue grew 110% this quarter, to $107.6bn. The bears say it is a capex cycle with no payback. Grant them the factory. Then look at the market it is being built to supply.
21.1bn connected devices were in use at the last count, 21.9bn forecast for 2026. 5.1m industrial robots in service. 143m AI-capable PCs a year, 52% of shipments.
And the silicon inside them is still trivial. Gartner puts automotive content at $860 per vehicle in 2024, rising to $1,902 by 2034. A fridge, a cart, a trailer, a boat — a few dollars each turns it into a customer of inference.
⚠️ The honest part: January carries 14 weeks, not 13. Per week, revenue grows 4.24% and gross profit per week is flat to slightly lower. Operating profit per week falls about 2%, and is negative across the whole guided range.
That is what an infancy looks like: enormous capital, visible strain, and a customer base that has not arrived yet because the product is only now getting cheap enough to fit inside it.
The data centre is the factory. The devices are the market. We have priced the factory.
macrotape.io/analysis/2026-1…
3
📈 October's inflation rate is not a number. It is a 26bp band — and the real yield did all the work.
The October 2026 CPI index printed 337.476. October 2025 is absent from the file. No denominator, no 12-month rate.
🔧 The two windows that survive:
• vs Sep 2025: 4.08% over 13 months
• vs Nov 2025: 3.82% over 11 months
The true rate sits between them. 26 basis points of statistical fog. The last computable print was Sep, at 3.72%. Core is a 19bp band: 2.68% to 2.87%.
This is a collection gap, not a revision. Payrolls kept October 2025 at 158,408k. CPI, core and unemployment all skip it. A rebenchmark would move everything; a factor change would restate levels, not delete a month.
🧩 Now the price. Over the four sessions to 13 Nov:
• 10yr nominal 4.86% → 5.15% (+29bp)
• real yield 2.63% → 2.91% (+28bp)
• breakeven 2.23% → 2.24% (+1bp)
96.6% of the move landed on the leg that needs no statistical print. The breakeven — the market's own 10yr inflation forecast — is at 2.20%, the 215th lowest of 218 sessions this year.
The real yield is now 56.5% of the nominal 10yr, the highest share of 2026, from 43.2% on 9 Mar. 2.91% is the 5th highest print of 218 sessions.
Two readings, two trades. Mechanically: with the anchor an interval, a repricing that wants to be about inflation has nothing to grip. Structurally: the real leg is where duration and fiscal risk have priced all year, so a move landing entirely there says this was never an inflation story.
📌 Where the answer arrives: CPI for November lands 10 Dec with a real base, because Nov 2025 exists. A point instead of a range.
Wrong above a breakeven of 2.34% before 10 Dec with the real yield flat. Or a real yield back under 2.63% with the breakeven still under 2.24%. Or a Nov CPI above 4.08%, which would put October outside the band and make the fence too narrow.
Full piece, with the charts: macrotape.io/analysis/2026-1…
Plain English: macrotape.io/posts/2026-11-1…
🏦 Reserves built $30.5bn in the week to 11 Nov, to $2,960.6bn. The Treasury's chequebook fell $24.1bn over the same week, so the one-for-one model said $2,954.2bn. We got $2,960.6bn. The residual flipped to +$6.4bn from −$43.1bn.
That is the second consecutive weekly build — the first back-to-back since 30 Sep and 7 Oct. It is real repair. It is not a policy turn.
⚠️ Over the full fortnight, $103.5bn left the chequebook and $66.8bn reached reserves. That leaves $36.7bn short — 35.4% of the cheque — that never arrived.
🔧 Secured money fixed at the administered rate on two consecutive sessions, on volumes above $3tn. SOFR 3.90% against IORB 3.90%: spread 0bp on 12 and 13 Nov. The collateral premium is gone, and no discount has replaced it.
📉 Credit gave back all of November's widening — high yield 3.23% from 3.48% — while still sitting 63bp above its August low. Equities sold off 1.9% into that combination. Funding is not what is being priced.
Invalid below a $2,930.1bn reserve print on 18 Nov, with the chequebook already rebuilt $12.5bn into 13 Nov.
macrotape.io/posts/2026-11-1…
Fancy less macro-slang?
Here is the English version :)
macrotape.io/posts/2026-11-1…
1
📈 The 20yr is the most expensive bond on the curve to sell, per dollar raised. It is also the smallest print the Treasury makes.
The 2026 coupon record, 66 sales from 12 Jan to 8 Oct: divide each auction tail by the dollars raised, and the ranking stops looking like a maturity ladder.
🔧 Tail in basis points per $1bn offered:
• 2yr 0.051
• 3yr 0.044
• 5yr 0.027
• 7yr 0.005
• 10yr 0.006
• 30yr 0.021
• 20yr 0.108
That is 5.1x the 30yr and 20x the 10yr. Per dollar, the market charges 5 times as much to take a bond 10 years shorter.
The 20yr is also the highest yield on the curve: 5.52% against 5.47% on the 13 Nov par close. A 5bp inversion, 217th of 218 sessions this year against a 0bp median. Unbroken for 45 sessions from 10 Sep.
🧩 And it is the smallest sale: $14bn average against $23bn at the 30yr. Dealers took 11.6% of the 20yr's private book against 9.6% of the 30yr's. On the worst 20yr of the year, 15 Sep, they took 16.7% and indirect bidders 52.1% — the lowest indirect share of any 20yr sale.
So this is not a supply price. Coupons settled $815bn in the last 90 days, down $419bn, or 34%, while bills carried 87.6% of $6,557bn gross. The 10yr term premium fell 21.1bp to 0.8738% in the month to 6 Nov.
No new duration supply, and the average long-end premium is falling — while the 20yr's premium over its neighbours is the widest of the year. The toll is being set by the balance sheet of the firm that has to warehouse a $14bn print no client mandate asks for. That is why the fee per dollar is highest exactly where the print is smallest.
⚠️ One measurement cuts against this: correlate the 20yr tail with the 20s30s spread on the 9 sales and you get −0.22. Nothing. On 9 observations that refuses the easy version of the argument, so the claim is about standing conditions, not about any single auction.
📌 Next test: the 20yr cadence has been third week of the month, Tuesday or Wednesday. The next one lands in the third week of November — 2 days away.
Wrong if 20s30s closes below zero for 3 consecutive sessions.
Full piece, with the charts: macrotape.io/analysis/2026-1…
1
📉 The 2s10s corridor compressed 7.5bp last week, to 27.4 from 34.9.
The short end did every bit of it.
🔧 On the 13 Nov marks:
• 2yr: 4.51% → 4.88% (+37.0bp)
• 10yr: 4.859% → 5.154% (+29.5bp)
• 3-month bill: +13.0bp
The 2yr rose 7.5bp more than the 10yr. That difference is the whole compression.
📊 The composition is the finding. Of the 10yr's 29.5bp, 28.5 was real yield and 1.0 was inflation compensation. Breakeven 2.24%, up from 2.23%.
Last week was the mirror: the 10yr fell 7.0bp and the breakeven fell 11.0bp, so the entire move came out of inflation. Two weeks, two repricings of the price of money, and inflation expectations have sat at 2.2% through both.
Gold confirms it this week: $3,842, down 6.5%. A metal that pays nothing falls when the real yield rises.
💳 Credit did not confirm anything. High-yield spreads 3.20% on 12 Nov from 3.48% on 5 Nov — a full retracement of the 2026 high. Investment grade 0.89% from 0.94%.
Equities added 2.5% into a 28.5bp real-yield rise. VIX fell to 19.30.
So the curve is pricing a higher real cost of money. Credit and equities are pricing nothing at all.
🚨 Wrong if the 2s10s widens back through 44.0bp — that makes 13 Nov a belly supply week, not a repricing of where policy ends up. Also dead if the breakeven clears 2.34% with the 10yr at 5.154%: that is the debasement read, not this one.
One of three markets is doing the work. Which one is early is the trade.
macrotape.io/posts/2026-11-1…
📈 The 10yr term premium fell 21.1bp in a month. The 1-year-ahead forward rose 38bp in 4 sessions.
Those 2 numbers are the same election, priced in 2 different places.
🔧 On the 12 Nov par close:
• 1y1y: 4.69% → 5.07% (+38bp)
• 5y5y: 5.08% → 5.25% (+17bp)
• 1y1y minus 5y5y: −39bp → −18bp
−18 is the 9th narrowest print of 217 sessions this desk holds. Median is −59. Every one of the other 8 narrow prints falls between 16 and 28 Sep.
Do the arithmetic on the shape. A repricing driven by duration supply concentrates at the back, where the coupon calendar lives. This one did the opposite: 2.7bp of near-horizon move for every 1 at the far end.
The long end went the other way. ACM 10yr term premium 1.0844% on 6 Oct → 0.8738% on 6 Nov. Still the 78th percentile of 2026, so a retracement of a 6-week build, not a collapse.
Mechanism: the 3 Nov result took the Senate and is projected to take the House. A majority short of 60 votes cannot pass a budget resolution that adds duration supply. The 4 Nov refunding held the 3yr, 10yr and 30yr sizes flat for a 4th consecutive quarter. Fixed supply, falling premium.
What survives is the reaction function. The Fed sets the policy rate whoever holds the House, so the variable that remains is who runs the institution and how fast a decision reverses. That prices at 12–24 months, not at 30 years.
One number complicates it: the 20yr closed 5bp over the 30yr, only the 11th time in 217 sessions. A market paying more for year 20 than year 30 has not stopped worrying about supply. It has stopped worrying about the last 10 years of it.
Breakeven held at 2.24% on both 5 and 12 Nov. The premium that fell was not an inflation premium.
Falsifier: 1y1y minus 5y5y back at −45bp on a daily par close inside 10 sessions.
Which borrower are you — the one rolling 12-month paper, or the one buying 30-year duration?
macrotape.io/analysis/2026-1…
The overnight market cleared its premium. The two-year didn't care.
SOFR fixed 3.89% on 10 November against a 3.90% administered rate — minus 1bp, and the third consecutive negative fixing after 10 positive ones. The Treasury did it: its account fell $24.1bn to $912.1bn, reserves rose $30.5bn to $2,960.6bn in the week to 11 November. $24.1bn out of the chequebook, $30.5bn into the bank base, 127% pass-through.
In the same window the 2yr par yield went 4.46% → 4.86%. +40bp in 4 steps, the largest such move in the 217 par sessions of 2026. 2s10s flattened 38.0bp → 24.0bp.
That is the whole point 🚨 Overnight cash is a policy variable the state can move with a chequebook. Two-year money is a claim on 12 policy decisions, and the chequebook has no price for those.
Every basis point of the 10yr move was real: nominal +26bp, breakeven unchanged at 2.24%, real yield 2.597% → 2.863%. Not one bp was charged for inflation. All of it was charged for the money itself.
Credit tightened 12bp and still cost 14bp more all-in — 3.30% high yield over a 5.10% 10yr is 8.40%. Borrowers paying more for less risk premium is the one combination that cannot persist.
Invalidation: the separation read dies if the 2yr retraces the 40bp while SOFR holds at or below 3.90%. That would make the front end a positioning event the funding market never confirmed.
If SOFR prints 3.89% again at the next fixing, where does the 2yr sit — still above 4.86%, or back under 4.60%?
macrotape.io/posts/2026-11-1…
📈 The 1yr Treasury rose 28bp in 3 sessions. The 30yr rose 10. The gap between them is now the narrowest it has been all year.
🔧 On the 10 Nov par close:
• 1yr: 4.28% → 4.56% (+28bp, 18bp of it in one session)
• 30yr: 5.25% → 5.35% (+10bp, 9bp in one session)
• 30yr minus 1yr: 88bp → 79bp
79bp is the lowest print in the 216 sessions this desk holds. Median is 117. Nothing sits below it, and one other session, 27 Oct, touched 80.
The 2s10s did the same thing, 32bp → 25bp, second narrowest of the 216 against a 47 median.
A curve has 2 jobs: price the policy rate for the next 2 years, and price the risk of lending for 30. When the front reprices 3x the back, the political fact landed on the first job and skipped the second.
Here is the part that is checkable and almost nobody checked. Supply did not cause it.
Gross coupon offering has sat inside a 360bn–371bn band every complete month of 2026. Nine months, 11 dollars of range. Bills carried 85.8% of gross settlement Jan–Sep. Coupons settled 106.1bn less across the last 90 days than the prior 90.
Less duration was sold and 12-month money got 28bp more expensive. That is not a quantity story.
It is a price story, and the price that moved is the one nobody watches. The ACM 10yr term premium fell 6.2bp to 0.8738% between 22 Oct and 6 Nov. The long end stopped pricing supply, so the repricing had to happen where the funding actually clears.
Now the number that reaches a balance sheet instead of a screen.
IG OAS: 0.91% (9 Nov), 3bp off its year high. HY OAS: 3.34%, 14bp inside its own.
The spread did not move. So the 28bp went straight into all-in cost:
• 12-month IG borrower: 5.19% → 5.47%
• 30-year IG borrower: 6.26%
Twenty-nine extra years of duration buys 79bp. That is the whole trade in one line.
Send this to the person who reads a bear steepener into every selloff. The curve flattened, and it flattened from the front.
If you fund at 12 months: how many bp of your cost moved this week with no change to your credit?
Wrong at 104bp on 30yr minus 1yr inside 10 sessions — the 25th percentile of 216. Re-widening requires either the long end selling off or the 1yr rallying back, and either one kills the claim that this was a 12-month event.
macrotape.io/analysis/2026-1…
10
📈 Brent fell 21.07% in 8 sessions. 2.95 of those points were a contract change, not a price move.
🔧 The continuous front month rolled off December onto January on 2 November. January traded 4.21 below it.
Same 8 sessions, matched contracts:
• Continuous front: 117.64 → 92.85 = −21.07%
• Fixed Jan27: 113.43 → 92.88 = −18.12%
The overlap is the tell. From 2 November both series print the identical number, which is the session the series adopted the January contract.
The correction reallocates the move, it does not undo it. Every tenor lost ground and the loss decays down the curve:
• Jan27 −18.12%
• Dec27 −11.28%
• Dec28 −7.85%
That shape is the point. A uniform fall is a demand shock — less consumption everywhere. A front-loaded fall is a dated event: mark down the barrel you might lose next month, leave the one 2 years out alone.
Here is what the flat price cannot tell you. Brent Jan27 still carries 18.73 dollars over Dec28, the 92nd percentile of 502 sessions. 41 sessions closed higher. The 30 Oct print of 34.47 was the high of the entire 2-year sample.
And the insurance refused the de-risking. OVX per dollar of Jan27 barrel went 0.5156 → 0.5804, up 12.57%, while the barrel fell 18.12%.
A market that reprices lower because the threat passed gets cheaper to insure faster than it falls. This one didn't.
Across 496 overlapping 5-session windows in 2 years, 1 combined a front-month fall of 15%+ with a rise in insurance per dollar. It ended 4 November.
Send this to the person who screenshots the Brent drawdown chart into a risk pack.
If you run crude exposure: what would your book look like with the 24-month spread at 10 dollars instead of 18.73?
Wrong below 10.00 on Jan27−Dec28 with insurance per dollar under 0.5156.
macrotape.io/analysis/2026-1…
The chequebook is no longer a forecast. That is the finding.
Reserves rose $36.3bn to $2,930.1bn as of 4 Nov — the first weekly build since 7 October, and the end of a $128.4bn two-week drain. Direction right.
Size wrong. The Treasury's general account paid out $79.4bn across those 7 days. $2,893.8bn plus $79.4bn points to $2,973.2bn. The print was $2,930.1bn. $43.1bn never arrived.
Second consecutive miss, same direction, and a larger share of the cheque: 54.3% against 48.8% the week before. A residual that grows as a share of the move is not a rounding error. It is a second flow.
Nothing visible absorbed it. Reverse repo take-up was $0.252bn on 9 Nov — functionally zero. Last week's $5.087bn print that looked like a cushion was the month-end calendar, gone in a session.
What the mechanism did deliver is the price. Secured overnight money fixed 3.89% against a 3.90% administered rate on 6 Nov: minus 1 basis point, ending 10 consecutive positive sessions. The collateral premium has retired. That is the cleanest confirmation this plumbing gives.
So the model changes status. It is a bound on the sign of the next move, not an estimate of its size.
⚠️ Wednesday's reserve print is the test. The account has already paid out a further $25.9bn since the reserve date, pointing to a third consecutive build. A print below $2,930.1bn against that payout makes the residual structural rather than a timing difference.
2s10s 32.0bp. 10yr 4.88%. 30yr 5.26%.
macrotape.io/posts/2026-11-1…
**February's Auction Showed 50% Foreign Demand. The Fed Wrote 14 Points of It.**
**The call:** The demand read off this week's 10yr and 30yr is a denominator, not a crowd — February carried a 28.3% SOMA add-on that printed the indirect share at 50.1% on total accepted against 64.3% on the 42bn the public actually had to buy. The two shares are linked by an identity, not a correlation: it holds to 0.02pp on all 75 nominal coupon auctions this year, with a −0.58 slope on the headline and −0.06 on private take.
**The falsifier:** A 30yr on 12 November clearing more than 2.0bp over its same-day par close with primary dealers above 15% of private take; or a 10s30s close above 63.0bp, the year's 90th percentile. No 2026 30yr has put dealers above 14.7% of private take.
**The instrument:** 10s30s at 40.0bp — the 18th percentile of 214 sessions. Levelled off the 6 November par close: 2Y 4.51 / 10Y 4.86 / 30Y 5.26 / 2s10s 35bp.
Also on the piece: this corrects our own 2 November concession ladder, whose bidder shares were all computed against total accepted. Same bidders, 5 to 8 points more real money than we printed.
Article: macrotape.io/analysis/2026-1…
The 10yr yield fell 7.0 basis points last week. All 7.0 of it, and 4.0 more, came out of inflation compensation.
The breakeven is 2.23% as of 6 November, from 2.34% on 30 October — 11.0 basis points, and the lowest print since 5 August. Meanwhile the real 10yr rose 4.0 basis points to 2.629%.
⚠️ That is not the Fed being priced easier. It is the bond market buying less inflation insurance and charging more real yield for the privilege. A policy-easing rally runs through the real leg; this one ran entirely through the inflation leg.
The 2s10s tells the same story from the other side. 34.9 basis points on the 6 November marks, from 36.9, and the long end did all the compressing: -7.0 on the 10yr against -5.0 on the 2yr.
And the 3-month bill went the other way, +3.0 to 4.19% — the only point on the curve to rise. The front of the front end is saying the near-term cash rate is not moving.
Credit is the disagreement worth watching. High-yield OAS at 3.48% on 5 November, the 2026 high and the widest since 9 May 2025, from 3.26% on 29 October. Twenty-two basis points wider in a week while breakevens collapsed.
The curve is pricing less inflation. The credit market is pricing more default. Those are not the same trade, and they rarely sit at the same address. Credit spreads widen before defaults arrive, not after — which makes the credit leg the earlier warning of the two.
📉 Invalidation, in one line: 2s10s back above 44.0 basis points.
macrotape.io/posts/2026-11-0…
📈 The vote moved the 2yr 12 basis points. It moved the 30yr 3.
One session, 5 November, par closes:
• 2yr 4.58% → 4.46% (−12bp)
• 10yr 4.90% → 4.84% (−6bp)
• 30yr 5.28% → 5.25% (−3bp)
The 2s30s — the gap between them, the plainest measure of curve steepness — went 70bp → 79bp. A 9bp steepening, the 4th largest in 212 sessions this year.
Everything fell. Nothing fell more than the tenor in front of it. That shape has a name: a bull steepener. It is what a repricing of the policy path looks like when it is NOT accompanied by a repricing of fiscal risk.
Which is what divided government should produce. One party cannot pass a budget resolution, so the deficit path stops being the live variable, so the term premium has nothing new to price. The Fed sets the policy rate whoever holds the House — and the surviving political question is who runs the Fed, which is a 2-year question.
Then the portfolio consequence, same tape:
• S&P −4.36%, largest one-day fall in the file
• TLT +0.21%
• Hedge beta −0.048
Strongest hedge in the same 5 down sessions: −1.545 on 22 October, when the S&P fell 1.21% and TLT rose 1.87%. The long bond protected you then. On 5 November it did not.
The instrument is the 2s30s at 79bp, not the 10yr at 4.84%. A 10yr that fell 6bp tells you nothing about which channel moved.
Invalidation: 2s30s back below 70bp with the 30yr leading the session. That restores the long end as the fiscal instrument and makes 5 November a single-session artefact. Or a hedge beta back at −1.545 on a 1% equity decline.
If you get one thing from this: a bond rally is not a bond rally. The front end prices the Fed, the back end prices the Treasury. This week only one of them had news.
Full piece, 4 charts: macrotape.io/analysis/2026-1…
3
🧪 Credit & Stress fell 0.371 to +0.350, the largest single-session drop in the 43 snapshots this desk holds. 0.326 of it — 88% — came from two legs: the VIX and the MOVE.
VIX 16.61 from 20.83, its largest one-day fall in the sample, with the S&P up 1.58%. MOVE 93.88 from 99.97. Insurance got cheaper; the debt did not.
The spreads went the other way over the week: high-yield OAS 3.40% (3 Nov) from 3.14% on 28 Oct, +26bp, with 19 of it excess spread. Investment grade 0.93%, +7bp and one basis point off its 1yr high. The all-in cost of high-yield debt rose 10.9bp on the week while investment grade's fell 8.1.
⚠️ Confirmed if VIX holds below 18 and MOVE below 95 for 3 sessions with HY OAS at or under 3.40% and IG at or under 0.93%. The migration case — better credits widening first — needs IG at 0.95% while HY holds above 3.40%; it came within 1bp twice in 5 sessions.
macrotape.io/posts/2026-11-0…
8
📈 Investment grade now costs more than the government's own duration risk.
On 30 October the IG index demanded 0.94% to lend to a corporation. The New York Fed's estimate of the 10yr term premium — what the government pays for borrowing long instead of rolling 3-month bills — was 0.9089%.
The gap: +3.11 basis points. The first positive print since 28 July.
Before that, 65 consecutive sessions in which corporate default risk was priced cheaper than sovereign interest-rate risk. At the 7 October low the gap was −26.63bp.
What moved it was not credit. Over the same 3 weeks:
• 10yr par 5.28% → 4.94%, −34bp
• 10yr real yield −32bp of that
• ACM term premium −17.7bp
• IG OAS +12bp, HY OAS +35bp
More than half the bond rally was the price of duration risk falling. Credit went the other way — into a 34bp rally, spreads widened.
The mechanism is the funding mix, not the election. Gross Treasury settlements, last 90 days vs prior: $7,931.8bn against $7,936.1bn. Flat.
Underneath: bills +$256.5bn, coupons −$260.8bn. Bill share 84.5% → 87.8%.
A quarter of a trillion rotated out of coupon supply while total supply stood still. A sweep cannot lift the term premium if there is no extra duration to absorb — so the premium fell with the rally, and the corporate premium crossed above it almost by default.
The instrument is the gap, not the 10yr yield.
All-in cost is the part a treasurer sees. HY all-in 8.37% → 8.38%, +1bp: the 34bp rally did nothing for the weakest borrower. IG all-in 6.10% → 5.88%, −22bp — the safest borrower got the entire gift.
That asymmetry is a transfer from risky borrowers to safe ones, and it is what a late-cycle credit market looks like from the liability side.
Invalidation: IG OAS back under 0.89% with term premium above 0.9389% — gap under −5bp, and 30 October was a month-end marking artefact. Or term premium above 1.0965%, the 2026 high, with the gap still positive: the fiscal path did reach duration and the supply mechanism is wrong.
Full piece, 4 charts: macrotape.io/analysis/2026-1…
19
🛢️ Brent fell 10.4%. The waterborne basis fell 25.5%.
January Brent 113.43 → 101.68 in 3 sessions. January WTI 102.21 → 93.32. The gap between them — the price of a barrel that has to cross water against one that sits in Oklahoma — went 11.22 → 8.36.
That is 2.86 of premium, the 2nd largest 3-session compression of 126 windows. On a 11.22 base it is 25.5% of the premium gone, against 10.4% of the flat price.
The carry agrees. Jan27 minus Dec28 lost 9.56, 34.47 → 24.91: the largest 3-session flattening in 6 months. The front is shedding a specific risk the back does not carry. A surplus would flatten the curve from the back.
The market that earns the premium did not follow. Equal-weight FRO/STNG/DHT/INSW/TNK sits at the 95.8th percentile of 2 years, 0.14% higher over the last 2 sessions, 10.97% below its 13 October peak.
Mechanism: divided government cannot legislate energy policy. It cannot pass a sanctions statute, appropriate an SPR programme, or legislate a corridor open. What remains is discretionary — enforcement, releases — and both land on the waterborne barrel. So the political fact reaches the basis, not the flat price.
OVX 61.69, 85.2nd percentile, down 1.00 on the session the basis compressed 1.14. Vol stays bid while the basis clears: the market will stop paying for the route but not sell the insurance.
Falsifier: Jan27 Brent-WTI back above 10.65 with the tanker index still above 259.83. Then the 2.86 was a squeeze and the premium never left.
macrotape.io/analysis/2026-1…
🪙 The sweep was the excuse, not the trade.
Democrats took the Senate and are projected to take the House. Nasdaq futures peaked at 28,512.00 at 23:30 UTC, printed 27,615.75 at 03:00 — a 2.54% hole — and are back at 28,361.00. The S&P contract did the same, shallower.
Gold is the tell: a 0.22% dip at 00:30, then 2.70% up to 4,048.80. An index loses 2.5% and gold does not bid — that is mechanical selling, not fear.
Bitcoin fell 2.66% to 75,615.52, recovered 3.56% to 78,304.91 and is 0.80% above the pre-result print. The dollar is the discrepancy: down only 0.45%. A sweep that markets believed would widen the deficit should move the currency more.
⚠️ Invalidation: Nasdaq back below 27,615.75 with gold under 4,000 — then the recovery was short-covering inside a de-risking event.
macrotape.io/analysis/2026-1…
3
📉 The corridor was 29.8 basis points on 23 October, the tightest since 25 September. By Friday's close it was 36.9. Seven days, 7.1 basis points, and the engine that did the work reversed.
📐 The 2yr fell 13 basis points to 4.56% and the 10yr fell 5.9 to 4.929%. The front end supplied the whole widening and the long end handed part of it back: 30yr down 2bp, the 3-month bill down 6. That is the ordering of a policy repricing, not a term-premium move. Last Monday's column read the opposite, with the long end doing the work and the front end sitting still.
⚠️ The administered rate never moved. IORB 3.90%, EFFR 3.88%. So the front end is pricing the next move, not recording one.
🔍 Funding says the other thing. Reserves fell $68.8bn in the week to 28 October to $2,893.8bn, the lowest since early December, while SOFR fixed 4 basis points above IORB. Easier money is being priced into a market that is charging more for cash.
📊 Credit is the only market agreeing: high-yield OAS tightened 14 basis points into a 3.5% Friday fall in the Nasdaq 100.
Invalidation: 2s10s back below 29.8.
macrotape.io/posts/2026-11-0…










