A record high on the narrowest breadth since 1990, and credit widening with nothing to offset it
The regime read — bottom line first
Macro Tide, sixth edition. A weekly reading of the macro regime, built from public data, that sizes a nine-sleeve portfolio by conviction and cuts exposure when funding, credit or the trend break. Three $100,000 books are marked to the latest close and logged each week — Macro Tide Core, Macro Tide Spring and Macro Tide Harbor — together with the prior model, kept running on its earlier rules, and an S&P 500 benchmark.
This edition covers three weeks rather than one, and is current to the Friday 9 October close. No note was published between 18 September and today: the author was travelling, and the discipline of this letter is to say so rather than to backfill commentary nobody wrote at the time. The macro series were rebuilt from source and are complete; the sleeve prices were re-fetched and checked against the exchanges to the cent.
The gauge barely moved. The things underneath it moved a great deal. The regime is still Goldilocks and conviction is −0.00 — flat to three decimals. But high-yield spreads widened forty-five basis points with the VIX under fifteen, the ten-year reached a level at which this model is willing to own duration for the first time, and the market set a record high on the thinnest participation in thirty-five years. None of those three moved the gauge. All three are worth your attention.
This edition leans forward rather than back. The week just ended is settled; what matters now is bank earnings, Wednesday's inflation print, and two dated events — the Federal Reserve in seventeen days and the midterm elections in twenty-three — that several people we read think will land on the long end of the bond market and the dollar. The last two sections say what we would act on and what we are watching.
The week that was
Three weeks passed without a note, and the index barely noticed. The S&P 500 closed Friday at 7,811.54, about 2% above where this letter last appeared on 18 September. Nothing in the regime broke, no guardrail armed, and the gauge finished roughly where it began: Goldilocks, conviction −0.00. September's month-end completed the move out of Reflation and October has held it. For a letter whose purpose is to say when the weather changes, the honest headline is that it did not.
But three things moved underneath, and none of them moved the gauge.
1. Credit widened, and this time there is nothing offsetting it.
High-yield spreads went from 2.70% to 3.15% — forty-five basis points — with the VIX under fifteen the whole way and repo never leaving the comfortable side of the floor. Investment-grade Baa moved only eight basis points over the same stretch. Eight against forty-five: whatever this is, it is happening at the weaker end of the credit market and not everywhere.
Three people we read independently saw the same thing from different angles. Mark Meldrum has triple-C spreads widening another 37 basis points last week to 12.52, against a cycle high of 12.89 set in September 2023, with triple-C yields at 17.32% — a level he calls simply unaffordable for companies already under stress. Dan Niles points at the banks: the KBWB bank ETF is down 12.1% since its 14 August peak while the S&P is up 0.3% over the same period. Torsten Slok at Apollo notes credit default swap spreads widening for AI companies specifically.
Meldrum supplies the mechanism, and it is the part worth understanding. In an ordinary cycle these two things move against each other. When the economy weakens, credit spreads widen but interest rates fall, and the fall in rates partly offsets the widening — so the all-in yield a borrower pays rises less than the spread alone suggests. That offset is what normally stops a credit wobble becoming a credit event.
Right now both are rising at once. Spreads are widening and the underlying interest rate is going up, so the total cost to the weakest borrowers is rising faster than either on its own. Unless the long end of the bond market rallies — yields falling, which would restore the offset — wider spreads will simply keep being met with higher yields and nothing will cushion them.
And Jeffrey Gundlach supplies the detail that changes what this is. Split the credit market in two and the widening is almost entirely one thing. Investment-grade excluding artificial intelligence borrowers has not widened at all. AI investment grade is out roughly 75 basis points over the past year. In high yield the non-AI market is close to its tights, while AI high yield is out by more than 150 basis points and probably more than 200 by now.
So this is not a general credit event. It is the market repricing one set of borrowers — the companies financing the data-centre build — and the average figures hide it. Gundlach's image for the whole market is a hundred-year-old maple that looked healthy until a branch came off and showed it was hollow inside.
Credit is the closest of our three fast-cut legs to its trigger. It was in September, and it is closer now. It has not armed. We are watching it rather than acting on it, and we would rather say that plainly than dress it up as a call. What we now know is where to look: not at the average spread, but at whether the AI split starts pulling the non-AI market with it.
2. The long end reached a level where we will own duration — and not everyone agrees.
The ten-year closed at 5.24%, the thirty-year at 5.60%. Meldrum notes the thirty-year printed a cycle high of 5.67% on 7 October and the ten-year and seven-year set cycle highs on the 5th, before pulling back a basis point or two into the weekend.
Ten-year breakevens did not move. Still 2.33%. That is the detail most commentary skates over, and Meldrum's reading of it matches our own exactly: inflation expectations are well anchored, so none of this is the market changing its mind about inflation. All of it is real yields. The price of money went up; the expected path of prices did not.
We size long bonds on one question: how far is the ten-year above fair value, where fair value is the average of the German Bund yield and seven-year nominal GDP growth — a rule we take from Jeffrey Gundlach. Below a quarter-point premium we own no duration at all. Fair value is 4.82% against a ten-year of 5.24%, so the gap is +0.42 and long bonds enter the book at 3.16% for the first time.
That is a valuation judgement, and three people we respect would not make it.
- Dan Niles uses a simpler rule — the ten-year should roughly equal nominal GDP — and gets a different answer. Real second-quarter GDP of 2.2% plus core PCE at 3.0% gives 5.2%, against a ten-year at 5.25%. On his arithmetic duration is fairly priced, not cheap. He adds that Friday's ten-year auction cleared at 5.30% against 5.32% trading beforehand, with dealers taking only 2.5% — real demand at this level, which makes him more constructive on what sold off.
- Mark Meldrum is blunter: TLT is down 10.5% this year and "the dividend is not paying you to wait." He expects one more rate rise priced for December and more beyond.
- Torsten Slok has argued both sides within four days, which is itself informative. On 6 October he set out why rates might peak within the month — midterms possibly easing Middle East tensions and oil, governments releasing reserves before voters go to the polls, higher rates biting housing and autos, and AI credit spreads widening. On 9 October he argued the opposite pressure: because AI investment is largely insensitive to interest rates, it keeps crowding out the sectors that are sensitive, leaving the Fed with AI-driven inflation alongside rate-sensitive weakness — and therefore higher for longer.
Jim Bianco is the one who agrees with us, and he is worth more than the others precisely because of where he is coming from. He has been bearish on bonds since 2020 and has just turned. His reasoning is not a forecast: in 2020 the average yield was 2% against a duration of seven, so a 30 basis point move against you wiped out the coupon. Today the yield is about 5.5% and the duration 5.5, so yields have to move a full point against you before you lose money. The cushion is back. His rule is a scale, not a call: "if the bond market is offering me mid-5% yields, I want to start getting in. If it wants to offer me high 5% yields, I want to get in more. If it gets to six, I want to get in even more." That is what a gate like ours does mechanically.
He is explicit that this is not a bet on yields falling. The economy is strong and inflation is sticky, and he still expects yields could rise further. The point is that at 5.25% you are finally paid for the risk, which was not true at 4%.
We have taken the position and we are not hiding the disagreement. Our rule says the long end is cheap by 42 basis points against a fair value that includes the Bund; Niles's rule, which ignores the Bund, says it is fair. Both are defensible. Ours is the one written down in advance, which is why we follow it.
And this week Gundlach told us what his own version of our rule currently reads.
He set the rule out in full: a seven-year moving average of US nominal GDP, averaged with the German ten-year yield, which he says has a 0.93 R-squared to the US ten-year going back forty years. That is our rule, and he is its author. Then he gave the current number.
He reads fair value at about 4.95% and the gap at roughly 30 basis points. We read 4.77% and 47.
We checked rather than assumed, and the difference is not the German yield. DoubleLine's own weekly market update for the 9 October close — Gundlach's firm, same day, independent of us — puts the ten-year Bund at 3.48% against the 3.50% we are carrying, and the ten-year Treasury at 5.24%, which is ours exactly. Our inputs are right to two basis points. It is the nominal-GDP term: we use 6.04%, his implies about 6.40%. Ours is deliberately lagged by a quarter so that the historical testing cannot see data that was not published at the time. That is the right choice for a backtest. It is a questionable one right now, because Michael Howell makes the point that 2026 is the fastest nominal GDP growth in the United States, and in the world, since the mid-1980s. When nominal growth is accelerating that fast, an average measured through an older window reads too low — and a fair value that reads too low makes the long end look cheaper than it is.
The consequence is not small. On our gap the gate asks for 4.73 points of long bonds and the book holds 3.16% after the momentum overlay. On Gundlach's own current reading the gate would ask for 1.40 points, and the book would hold about 0.9% — roughly a third of what we own.
We are not changing the model in response to a podcast, and we have not. The lag is in the rules, the rules were set in advance, and the position was sized by them. But the author of the rule is reading his own indicator differently from the way we are reading it, for a reason we can identify, and you should know that before you read our duration position as his endorsement. It has gone to the research thread as a specified question rather than an adjustment.
3. Breadth is far worse than our own measure shows.
This is where our instruments and everyone else's diverge, and we would rather show that than smooth it over.
Our measure says the narrowing stopped. Three weeks ago the equal-weighted S&P trailed the cap-weighted index by 6.85% over six months. It has roughly halved, to −3.09%, and the twelve-month measure — the one the model actually reads — has gone from a −3.22% drag to −0.18%, which is effectively nothing.
Four outside measures say participation is the weakest in a generation.
- Meldrum: on the record high of 6 October, only about 30% of S&P components were above their 50-day moving average — the weakest participation on a record day since 1990. He also cites Bloomberg at 43% of components more than 20% below their 52-week highs, FactSet at 41%, and Nomura at 59% below all-time highs.
- Niles: 35% above the 50-day at week's end, improved from 25% the week before. The Magnificent 7 rose 1.6% on average while the Russell 2000 fell 0.9%.
- Bespoke Investment Group: the Russell 3000 is up about 3% in the second half while the average stock is down 4.4%.
- Meldrum again: more than half the components of the Russell 2000 are more than 20% below their 52-week high — over half the small-cap index is in a bear market.
Both readings are true, and they measure different things. Ours compares 500 large companies to each other; theirs measure how many stocks are participating at all, across the whole market. A narrow advance can halve the large-cap equal-weight gap while leaving most listed companies out of the move entirely. That is what has happened.
It does not change what we do, and this week we can finally say why with numbers rather than assertion. We asked the research thread whether this book has a structural weakness in narrow, mega-cap-led advances. The answer came back and it is the opposite of the worry: measured within rising months, the books do better when the advance is narrow than when it is broad — by roughly 3 to 11 percentage points a year across the three products and both test samples. The reason is in the design: we hold US equity through cap-weighted and mega-cap vehicles, never equal-weight, so we are long the narrowness. The one sleeve beating the index this quarter is mega-cap growth, and that is the design working rather than failing.
So narrow breadth is something we report because it is true and uncomfortable, not because it is a signal we trade. The first measurement of it, incidentally, came back saying we had a severe problem — a 20-point-a-year shortfall — and was wrong, because the condition being measured included "the market rose" and a book carrying half the index's risk trails a rising index by construction. Measured properly, the sign reverses. We publish that because we said we would.
The thread running through all three.
Put the pieces side by side and they are one story rather than three.
Bianco's framing is the clearest: there are now two stock markets inside the S&P. About fifty AI-related names — the mega-caps, the semiconductors, the equipment makers — are roughly 45 to 48% of the index. The other four hundred and fifty are the rest. Their correlation with each other is falling, so on a given day the index can drop because the AI half fell while the other half rose, and the headline will say investors are worried about growth.
Now lay Gundlach's credit split on top of exactly the same divide. Investment grade excluding AI has not widened at all; AI investment grade is out 75 basis points. Non-AI high yield is near its tights; AI high yield is out more than 150.
It is the same split, in two asset classes at once. Our breadth problem and our credit widening are not two separate warnings. They are one concentration, showing up in equities as narrow participation and in credit as a single sector repricing. Slok's observation that AI capex is insensitive to interest rates explains why it can keep going while everything rate-sensitive struggles, and why the Federal Reserve is stuck.
That is the thing to hold in mind over the next month. Not "is the market overvalued" but "is the AI complex still funding itself" — because that is the question both halves of the split are asking, and credit usually answers before equities do.
Where that leaves the books.
All four books were rebalanced onto the refreshed targets and are sitting on them. The 30 September month-end allocation, which was never struck while the letter was dark, has been computed from that month-end's own data and added to the record, dated as published on 10 October so the reconstruction is visible rather than slipped past you.
Action: hold. No sleeve moved more than half a point, the regime is unchanged, and the fast-cut is clear on all three legs.
What changed this week
- Allocation shifts (>3pt, or a sleeve entering or leaving): Gold -3.5pt, Long bonds +3.2pt (enters the book).
Sleeve moves — from % to %, and why
Target weights of the Macro Tide Core, last note → this note, every sleeve. A move is either the rules (a methodology change, dated in the changelog) or the regime (conviction, quadrant, a gate) — the why says which. Tilt figures are computed from this edition's z and this product's own tilt strength. The withdrawal dial (below the allocation table) sits on top of these and never changes them.
| Sleeve | Last note | This note | Why |
|---|---|---|---|
| US equity (SPY+MAGS) | 45% | 45% | — |
| Int'l developed (VEA) | 12% | 11% | momentum tilt z +0.95 → ×1.47 |
| Emerging mkts (EEM) | 13% | 12% | momentum tilt z +1.06 → ×1.53 |
| Diversified commodities (PDBC) | 14% | 14% | — |
| Gold (GLD) | 14% | 13% | momentum tilt z -0.53 → ×0.73 |
| Bitcoin (IBIT) | 2% | 2% | — |
| Managed futures (DBMF) | 0% | 0% | — |
| Long bonds (TLT) | 0% | 3% | momentum tilt z -0.57 → ×0.71 |
| Cash (USFR) | 0% | 0% | — |
Bold = moved ≥ 0.5pp. History accrues in targets_log.csv by edition.
Where we are — the 2×2
The liquidity tide
The six-cycle read
Each row is the CURRENT-WEEK standing — every cycle is recomputed on the latest data (not the frozen month-end): z-scores vs each series' ~10-yr norm for the composites (+ = above average), 13-week momentum for Liquidity, and the spread level for Credit. Last (18 September) is the reading in the previous note and Δ 3wk the move since; what changed underneath is in the sections below. Δ in this edition is measured from the 18 September close — 3 weeks, not one. Measuring reaches back that far because the week ending 2 October was never logged and the 25 September row is a reconstruction rather than a reading taken at the time, so it is not used as a baseline.
| Cycle | Last (18 September) | Reading | Δ 3wk | State |
|---|---|---|---|---|
| Growth | +0.45 | +0.48 | +0.03 | accelerating · tape: firm ▲ |
| Inflation | +0.13 | -0.06 | -0.19 | cooling · tape: mixed → |
| Liquidity (Fed) | -0.8% | -0.7% | +0.1pp | risk-off |
| Liquidity (Global) | -1.7% | -1.4% | +0.2pp | risk-off · Fed+ECB+BoJ (PBoC pending) |
| Treasury liq | — | +0.80 | — | injecting — Bessent's bridge |
| Monetary | -0.94 | -0.94 | +0.00 | tight |
| Positioning | -0.30 | -0.55 | -0.25 | crowded |
| Dollar | weak | firm | — | neutral |
| Credit (veto) | HY 2.70% | HY 3.15% | +0.45 | off |
Native readings (under the hood): Growth = 9/11 indicators above trend · Inflation = 4/8 up (broad underlying ~2.6%) · Liquidity = net $5.9T · Monetary = 2y 4.75%, real 2.87% · Positioning = VIX 15 · Dollar = -0.0% y/y (broad TWI).
How to read the scores. The z column counts standard deviations from that cycle's own ten-year average, so it says how UNUSUAL a reading is rather than what it is. Zero is normal, half a point is mildly above average, one point is unusual and two is rare. Leading indicators carry the most weight in the average. Growth, Inflation, Monetary and Positioning all lean on monthly data, so the tape flag beside Growth and Inflation is the fresher read: it comes from five days of market moves across asset classes and can run a month or two ahead of the slower score it either confirms or contradicts. Two rows work differently — Liquidity is thirteen-week momentum, not a z-score, and Credit is simply the spread level.
Treasury liq (Bessent's bridge): the Treasury injects liquidity via long-bond buybacks + bill-heavy issuance — swapping duration for near-cash, which eases financial conditions like stealth QE without the Fed cutting. +0.80 = injecting; a displayed read, not wired into conviction.
Dollar (how it's measured): the Fed's broad trade-weighted USD (DTWEXBGS, 26 currencies incl. China/Mexico/EM), -0.0% YoY and below its 12-mo trend. The narrower, euro-heavy DXY can read firmer — the softness is concentrated against trade/EM partners, which is exactly the debasement tilt (gold + EM).
Liquidity — current weekly read (Fed H.4.1), now wired into conviction: net Fed liquidity $5.87T, 13-week momentum -0.7% as of 2026-10-07 — vs the trailing monthly composite's -0.8%. The drain is steady (TGA rebuild into a drained RRP now pulls on bank reserves directly). Every cycle is now recomputed weekly on the latest data; the monthly composite remains the historical backbone the states are smoothed against.
Global liquidity (Howell). We add up the big central banks' balance sheets in dollars — Fed $6.7T, ECB $6.6T, BoJ $4.0T, $17.3T in all — and track whether that total is growing or shrinking over thirteen weeks. It is running at -1.4% (as of 2026-10-02; the global figures lag about a week), against -0.7% for the Fed on its own, because the ECB and BoJ are shrinking too. This feeds conviction alongside the Fed read, and the two agree this week. On our data it leads the S&P's next-quarter return by about thirteen weeks (roughly 0.6 correlation), so a shrinking tide is a headwind ahead rather than today — which is Howell's “liquidity peaked around the end of 2025” call. Note this is the slow, leading read: the fast-cut's funding test watches US money-market plumbing directly instead (SOFR against IORB). China's central bank, which Howell ties to gold, would be a fourth leg; there is no clean free data series for it yet.
Cross-asset tape
The market's leading edge vs the monthly composite — 5-day moves as of 2026-10-09.
Nowcast — Inflation mixed →, Growth firm ↑. The six-cycle inflation reading is recomputed weekly (now -0.06) and its momentum state still reads cooling; the cross-asset tape (5-day) agrees (oil, 10Y). This tape read is the leading edge, shown not wired into the sleeves: if it persists, the momentum state flips and conviction gets cut (inflation enters as −i_state at a 0.25 weight). Growth internals are noisier — one soft breadth print isn't a turn while the hard data (payrolls) is firm.
Color = direction of the 5-day move, not good/bad. Free data (Yahoo): UST yields, DXY & major FX, commodity ETFs, diesel (ULSD future — the real-economy barometer of the oil shock; diesel outpacing crude = the cost is running ahead of the headline), HY/IG credit, equal- vs cap-weight breadth, listed BDCs vs HY (the private-credit tell: BDCs lagging high-yield = the market pricing private-credit stress ahead of the quarterly marks).
Factor lens — 1D · WTD · QTD · YTD
Style factors vs the S&P 500, excess return in percentage points, as of 2026-10-09 (WTD from the 2026-10-02 close, QTD from 2026-09-30, YTD from 2025-12-31). z = this week's 5-day excess vs its own 1-year distribution. Shown, not wired into the sleeves.
| Style (ETF) | 1D | WTD | QTD | YTD | z (5d) |
|---|---|---|---|---|---|
| Momentum (MTUM/SPY) | +0.24 | -2.05 | -1.25 | +13.67 | -0.93 |
| Value (IWD/SPY) | +0.01 | +0.51 | +0.17 | +6.37 | +0.36 |
| Small-cap value (IWN/SPY) | -0.64 | -1.85 | -1.41 | +2.20 | -1.10 |
| Small-cap (IJR/SPY) | -0.43 | -1.86 | -0.98 | +0.28 | -1.05 |
| High dividend (VYM/SPY) | -0.28 | +0.30 | +0.20 | -3.56 | +0.28 |
| Growth (IWF/SPY) | +0.03 | -0.50 | +0.03 | -6.07 | -0.37 |
| Low-vol (USMV/SPY) | -0.09 | +1.20 | +1.08 | -7.40 | +0.94 |
S&P 500 itself: 1D +0.60% · WTD +1.16% · QTD +2.09% · YTD +14.17%.
Sleeve scoreboard — the nine sleeves vs the S&P
Absolute return and excess vs SPY over the same horizons — the model's own building blocks, watched the way the factors are.
| Sleeve (ETF) | 1D | WTD | QTD | YTD | vs SPY WTD | vs SPY QTD | vs SPY YTD |
|---|---|---|---|---|---|---|---|
| US equity (SPY) | +0.60% | +1.16% | +2.09% | +14.17% | +0.00 | +0.00 | +0.00 |
| Mega-cap growth (MAGS) | +1.01% | +1.63% | +2.90% | +11.84% | +0.47 | +0.81 | -2.33 |
| Momentum (MTUM) | +0.84% | -0.89% | +0.84% | +27.84% | -2.05 | -1.25 | +13.67 |
| Intl developed (VEA) | +0.70% | -1.10% | -0.37% | +12.61% | -2.26 | -2.46 | -1.56 |
| Emerging (EEM) | +1.06% | -1.29% | +0.01% | +22.10% | -2.45 | -2.08 | +7.93 |
| Commodities (PDBC) | +0.05% | +1.13% | +1.87% | +48.38% | -0.03 | -0.22 | +34.21 |
| Gold (GLD) | +1.57% | +1.17% | +0.98% | -2.96% | +0.01 | -1.11 | -17.13 |
| Bitcoin (IBIT) | +0.63% | -2.47% | -1.67% | -6.24% | -3.63 | -3.76 | -20.41 |
| Managed futures (DBMF) | +0.73% | +0.73% | +1.91% | +18.14% | -0.43 | -0.18 | +3.97 |
| Long bonds (TLT) | +0.14% | +0.65% | +0.26% | -10.53% | -0.51 | -1.83 | -24.70 |
| Cash (USFR) | +0.02% | +0.08% | +0.10% | +0.24% | -1.08 | -1.99 | -13.93 |
Styles. Year-to-date momentum leads (+13.7pp vs the S&P) and low-vol lags (-7.4pp); quarter-to-date low-vol leads (+1.1pp) and small-cap value lags (-1.4pp); this week low-vol (+1.2pp) over momentum (-2.0pp). The leader changes with the horizon — rotation inside the market rather than one persistent bet.
Value vs growth. Value is 12.4pp ahead of growth year-to-date — the reflation/late-cycle signature (higher yields, oil, a weaker dollar) that the inflation-tilted regime read implies; it is also why the equal-weight index has stopped underperforming.
Risk appetite. Low-vol -7.4pp YTD — a market paying for beta, not safety (risk-ON), which matches the positioning cycle's 'neutral, not fearful' read.
Momentum. +13.7pp YTD but -1.2pp QTD: the July crowding reversal, the multi-strategy deleveraging Burry described, returned the excess. The US-equity sleeve holds momentum only while conviction is positive, the condition designed for this case.
Sleeves, year-to-date. gold -3.0% (-17.1pp vs SPY); commodities +48.4% (+34.2pp vs SPY); EM +22.1% (+7.9pp vs SPY); intl developed +12.6% (-1.6pp vs SPY); mega-cap growth +11.8% (-2.3pp vs SPY); managed futures +18.1% (+4.0pp vs SPY); long bonds -10.5% (-24.7pp vs SPY); bitcoin -6.2% (-20.4pp vs SPY).
What it tells the model. Commodities are well ahead of the S&P while long bonds lag — the pattern an inflation tape makes. The real-asset sleeve is positioned for it. Long bonds are the exception: the tape says avoid them, while the fair-value rule has bought 3.2% of the book because the ten-year yields more than the rule says it should. Tape and valuation disagree here, and we are following valuation. Bitcoin is far behind gold. Both are meant to gain when money is being debased, but bitcoin also needs plentiful liquidity, and it is not confirming the move. The tilt keeps the position small.
5-day risk pairs: High-beta vs low-vol -2.22pp · Cyclicals vs defensives -1.05pp · Small vs large -2.08pp. Read: Low-vol is the leading style (+1.20pp vs the S&P); growth ahead of value by 1.00pp; factors lean risk-OFF; with a defensive bid (low-vol + quality bid up).
The Gundlach dashboard
Four of Jeffrey Gundlach's standing metrics, reproduced from free data (Sep-8-2026 webcast, "The Fed's Next Move"). Shown as context for the Monetary cycle and the duration gate — none is wired into conviction.
| Metric | Reading | His rule | What it says now |
|---|---|---|---|
| 10-year fair value: (Bund 3.50% + 7-yr avg nominal GDP growth 6.14%) ÷ 2 | 4.82% vs actual 5.24% | average of the two; R² 0.93 since 1986 | the 10-year is +42bp above the model — well away from the model (Bund as of 2026-10-09; GDP y/y last 6.8%, 2026-04) |
| 2-year vs fed funds | 2Y 4.80% − FF 3.88% = +92bp | the 2-year leads the Fed (2022: +237bp, the widest of his career) | the 2-year says the Fed is BEHIND — hike bias — the funds rate 'should' be ~100bp higher on his read |
| Copper/gold ratio (CPER/GLD) | 0.105 · +2.4% 3-mo · +24.3% 12-mo · 23th pct since 2018 | moves with the 10-year; ratio up = yields up | flat — no push either way |
| Shiller CAPE | 41.9 (mean 17.4, median 16.1, 1999 peak 44.2) | 10-yr forward real return vs CAPE, 1965–2015 | above 40 — in his 1965–2015 regression no 10-year forward REAL return from here was positive (multpl.com, 2026-10-12) |
| Import / export prices, y/y (his "purest" inflation) | imports +6.8% · exports +8.8% (2026-08) | unadjusted, no hedonics — "inflation is really running ~7%" on his Sep-8 read | elevated on both sides |
How we use it: the 10-year model and the 2-year spread inform the Monetary cycle's narrative and the duration gate (gate opens only when the monetary read turns); CAPE is valuation context — our own point-in-time test found valuation does not time entries, so it stays out of conviction; copper/gold is a tape cross-check.
The week in numbers
Last note vs this note for the gauge; the tape by day from the daily monitor's snapshots. What moved, in one table.
| Gauge | Last note | This note | Why |
|---|---|---|---|
| Conviction | +0.01 | -0.00 | +0.05 → +0.01 after the Saturday data audit: the corrected 2-year (4.34 typed → 4.56 on FRED) pushed monetary lower and the August CPI/PPI prints entered the gauge; positioning flat. Same neutral band, targets unchanged |
| Growth | +0.45 | +0.48 | hard data firm (payrolls, ISM services); the soft data flinched — Michigan 47.8 |
| Inflation | +0.13 | -0.06 | August core CPI (+0.3% m/m) and commodity PPI now in the gauge point-in-time: +0.08 → +0.14; oil pass-through continues |
| Fed net-liq, 13-wk | -0.8% | -0.7% | H.4.1: TGA drained $85bn — better, still ebbing; state stays risk-off (flips above +1%) |
| Monetary | -0.94 | -0.94 | −0.60 → −0.75: the 2-year is 4.56 (FRED 9/10), not the 4.34 the file had carried since 9/4 — 100bp over funds, Gundlach's 'Fed is behind' read is stronger than we showed |
| Positioning | -0.30 | -0.55 | −0.49 → −0.27: two-thirds is the new MOVE input (method), the rest a higher VIX (data); Friday's relief rally took a little back |
| Risk budget | 80% | 80% | rules: the neutral step moves 0.45 → 0.55 (changelog §11); conviction stays in the neutral band, so the Base model's budget is 55% × 1.35 ≈ 74% — unchanged by the data refresh |
| 10-yr vs Gundlach fair value | +0.19pp | +0.42pp | 10-yr 5.24% vs fair value 4.82%; TLT at 34% of its regime weight (scales in from +0.25pp) |
| Commodity gate | OFF | OFF | liquidity z +0.07, monetary z -0.43 vs −1 |
| Tape | Fri 10/9 | Fri 10/9 | Fri 10/9 | Read |
|---|---|---|---|---|
| 10-year (close) | 5.24% | 5.24% | 5.24% | repricing the policy path; 30-yr at its 2007 high |
| 2-year (close) | 4.80% | 4.80% | 4.80% | leads the Fed |
| USD/JPY | 158.2 | 158.2 | 158.2 | quiet — carry-unwind flag by history only |
| SOFR − IORB (spot) | -3bp | -3bp | -3bp | funding leg clear (21-day avg vs +5bp) |
| ON RRP | $0.3bn | $0.3bn | $0.3bn | drained |
| Fast-cut | HOLD | HOLD | HOLD | all three legs clear all week |
What we learned this week
The inputs that came in, and what the model did with each — adopted, tested and kept, tested and rejected, or recorded as commentary. Details in the model changelog.
• Our rule for owning long bonds disagrees with the man who wrote it, and we found out why (Sun). The rule is simple: long bonds are worth owning when the ten-year yields more than fair value, and fair value is the average of two numbers — how fast the economy has grown in cash terms over seven years, and the German ten-year. Gundlach built it, and it has tracked the US ten-year closely for forty years. He puts fair value at 4.95% today. We get 4.77%. Same rule, different answer, so one of the two inputs must differ. It is not the German yield: ours is 3.50% and his own firm published 3.48% for the same day. It is the growth number — we use 6.04%, he implies about 6.40%. Ours is deliberately a quarter out of date, because our backtest must never use a figure before the day it was published. That honesty costs us precisely when growth is speeding up, and Howell says 2026 is the fastest since the mid-1980s. What it does to the book: we see the ten-year as further above fair value than he does, so our rule asks for more long bonds — 4.73% of the book against roughly 1.4% on his reading. We hold 3.16% once the momentum overlay has had its say. This went to research as a question. Nothing in the model changed.
• The credit widening is almost entirely one sector (Gundlach). Investment grade excluding AI has not widened at all; AI investment grade is out ~75bp over a year. Non-AI high yield is near its tights; AI high yield is out more than 150bp. Our own +45bp on the high-yield index is an average concealing a split. The thing to watch is whether the AI leg starts pulling the rest with it.
• Both legs of the borrower's cost are rising at once (Meldrum). In an ordinary cycle spreads widen while rates fall, and the fall cushions the widening. Right now spreads and the underlying rate are both rising, so the all-in cost to the weakest borrowers compounds. Triple-C yields at 17.32% against a 12.89 cycle-high spread from September 2023. Unless the long end rallies, there is no shock absorber.
• Our breadth measure and everyone else's measure different things, and ours is the narrow one. RSP/SPY says the narrowing reversed (−0.18% at twelve months). Meldrum: 30% of the S&P above its 50-day on the 6 October record, weakest participation on a record day since 1990. Gundlach: 440 of the S&P 500 are more than 21% below their highs. Rosenberg: 75% of the S&P 1500 in correction. Bespoke: Russell 3000 +3% in H2 while the average stock is −4.4%. All true; ours compares large caps to each other, theirs count participation.
• Narrow advances are where this book does better, now measured rather than asserted (research, N1). Within rising months the three products beat their broad-advance excess by roughly 3 to 11 percentage points a year on both samples. The first measurement said the opposite because the condition included "the market rose" and a half-beta book trails a rising index by construction. N1 fired, no candidate designed.
• The rate debate has four defensible answers and we should say so. Niles: ten-year equals nominal GDP, 2.2% real plus 3.0% core PCE, so 5.2% against 5.25% is fair, not cheap. Slok on 6 October: rates may peak within a month. Slok on 9 October: AI capex is rate-insensitive, crowds out housing and autos, so higher for longer. Meldrum: TLT down 10.5% on the year, the dividend is not paying you to wait. Gundlach wants a 6-handle on the thirty-year and a 300bp real yield before he buys it.
• Breakevens have not moved, and that is the whole story of the move (Rosenberg, Meldrum, our own data). Ten-year breakevens at 2.33%, five-year-five-year forwards at 2.34%, flat for five years. Every basis point of the rise has been real yield. Rosenberg attributes it to the new Fed chair's reaction function rather than to debt or deficits, which were already there in February when the ten-year was 3.97%.
• A trap worth recording: thirty-year TIPS and nominals are effectively perfectly correlated (Gundlach). Switching long nominals for long TIPS to hedge inflation does not hedge anything; both fall together. Noted because it is the kind of thing that sounds prudent and is not.
Inflation cross-check · COVID-robustour standard read (-0.06) measures inflation against a decade whose average was inflated by the 2021–23 spike. On a COVID-robust basis (median/MAD) it reads +0.17, and broad underlying inflation still sits +0.58pp above the 2% target — the disinflation is off the peak, not below normal.
What moved under the hood
Which of the 19 sub-indicators shifted this week — only the movers (|Δz| ≥ 0.25).
No sub-indicator moved past the ±0.25 threshold this week — all 19 held.
Net: Growth +0.03 → +0.48 · Inflation -0.19 → -0.06 · nothing flipped a cycle.
The week ahead
This week decides whether the credit wobble is about a few weak borrowers or about the economy. Bank earnings run from Monday through Thursday and they arrive with the bank index already down 12.1% from its August peak while the market sits near a record. Wednesday brings the inflation print. Those two together will tell us more than the last three weeks did.
Bank earnings are the credit test, not an equity story. Dan Niles puts it directly: credit is the lifeblood of the economy, and the bank index falling 12.1% since 14 August while the S&P rose 0.3% is a divergence one side has to resolve. If banks sell off further from an already oversold level, the market is saying the widening we measured is about more than a handful of stressed borrowers. If they hold, it stays a weak-borrower story and our fast-cut leg stays where it is.
The long end is now a position, so we care which way it breaks. We own 3.16% in long bonds for the first time, on the judgement that the ten-year is cheap against fair value. Wednesday's inflation print goes straight at that, because breakevens have not moved and the entire rise in yields has been real rates. Two dated events sit behind it. The midterms on 3 November are, in Mark Meldrum's reading, a risk to the dollar and the long end specifically rather than to the short end, which is monetary policy and will not be changed by an election — and he expects the reaction to the result to matter more than the result. Then on 4 November comes the Treasury's quarterly funding announcement. David Rosenberg points at the precedent: in early November 2023 the Treasury cut issuance at the long end and flooded the system with bills, and the ten-year fell about 100 basis points into year-end. Nothing says they will do it again, but it is the one scheduled event that could move the long end quickly without the Federal Reserve doing anything.
What would make us wrong on duration. A position without a disconfirming condition is just an opinion, so here are ours. If nominal growth re-accelerates, fair value rises and the gap closes without yields falling at all — Niles's version of the rule, which tracks nominal GDP alone, already shows no gap, and Gundlach's own reading of our rule is 18 basis points above ours. If instead Torsten Slok's second argument is right and AI investment keeps the Federal Reserve tight regardless of what that does to housing and autos, the ten-year can sit above fair value for a long time while we hold a position that is right on valuation and wrong on timing. Meldrum's version is the blunt one: TLT is down 10.5% this year and the dividend is not paying you to wait.
Michael Howell's framework says where we are in the cycle, and it points the same way we do. His measure of global liquidity peaked in late 2025 and has been falling since, leading the real economy by roughly nine months, with money draining out of financial markets precisely because the real economy is strong — which is also his explanation for rising yields rather than a debt panic. On his phase map we are leaving speculation and entering turbulence, and turbulence means more bonds and less risk. He also names the two instruments he watches for stress, both of which are already in our tape: the MOVE index of bond volatility, and repo spreads. One caution he adds: price-to-earnings multiples are contracting while earnings are still accelerating, which is unusual, and he expects earnings to catch down rather than multiples to catch up.
What we are doing: nothing. Action is hold. The regime is Goldilocks, conviction is flat at minus zero, the fast-cut is clear on all three legs, and no sleeve moved more than half a point. The duration position was established this week and is sized by rule rather than by view; it needs no adjustment and will scale itself as the gap moves.
Three things that would change it, in order of likelihood.
- Credit arming the fast-cut. It is the closest leg, and it is the single thing most likely to move this book in the next month. It would cut risk across every sleeve rather than adjust one.
- The gap closing on its own. If fair value rises to meet the ten-year, the duration position shrinks without anyone deciding anything.
- A genuinely broad advance. Counter-intuitively this book does better when gains are narrow than when they are broad, which research measured this week. Small caps and the average stock participating is the condition in which we lag. Nobody is asking about it, which is usually when it is worth saying.
Scheduled releases that could move the gauge — consensus where available.
| Date | Release | Feeds | Consensus | Prior |
|---|---|---|---|---|
| Mon 10/12 | Bank earnings begin: JPMorgan, Citigroup, Wells Fargo, Goldman | Credit | — | the direct read on whether credit is deteriorating |
| Tue 10/13 | Bank of America, Morgan Stanley, State Street, BlackRock | Credit | — | — |
| Wed 10/14 | Consumer price index (Sep) | Inflation | breakevens anchored at 2.33% — a hot print tests that | — |
| Wed 10/14 | Charles Schwab, BNY, PNC, US Bank | Credit | — | — |
| Thu 10/15 | Producer prices (Sep) | Inflation | — | — |
| Thu 10/15 | Retail sales (Sep) | Growth | the consumer side; sentiment fell sharply in the Oct prelim | — |
| Thu 10/15 | Regional banks: Citizens, M&T, Truist, Travelers | Credit | — | — |
| Wed 10/28 | FOMC decision | Monetary | futures ~82% no change; the interest is the balance sheet | — |
Real-world check
Real-world check: where the desks agree with the reading, and the one place they do not. BlackRock's Global Weekly (14 September) stays pro-risk and keeps its overweights in US equities and the AI capital-spending theme, arguing that when higher yields reflect stronger investment, the resulting earnings can offset a higher cost of capital. That is the same growth-is-real judgment the gauge makes, and it is the model's stated blind spot read back from the other side. The disagreement is emerging markets: BlackRock upgraded EM equities to overweight in the same commentary, and the model cut its EM sleeve by four points this week. The cut is mechanical, the output of a relative-momentum tilt measured against the S&P 500, not a view on the asset class. If EM is where the desks are moving, the tilt will pick it up with a lag rather than lead it; that is what a rules-based sleeve does, and it is the cost of not overriding the rule. On the Fed, the Street moved further than the model did. Goldman Sachs told clients on 15 September that it expected the hike but not a successor; within a day of Wednesday's decision it moved to a second 25bp increase in October, citing the committee's push for a timelier return to target. The SEP supports that reading: sixteen of eighteen officials see at least one more increase this year. The gauge does not forecast the path; it reads the price of it. The monetary cycle tightened to −0.94 on a two-year yield at 4.76%, which is the market's own version of the Goldman revision. The hike passed through the funding market almost perfectly, which is what keeps the stop unarmed. Interest on reserves re-based to 3.90% and SOFR printed 3.85%: 23 of 25 basis points transmitted, no reserve scarcity. Fed net liquidity improved to −0.8% on thirteen-week momentum and the global tide to −1.7%. Both agree, so global liquidity enters the conviction score this week; both are still draining. The single reading that would change the assessment is a SOFR fix that holds above the rate on reserves, and nothing in this week's prints points to it. The counterarguments worth holding. First, the regional surveys: Empire at 7.6 and Philadelphia at 37.8 cannot both describe September, and the composite currently sides with the stronger of the two. Second, the yen: a currency that weakens on a Bank of Japan rate rise is the configuration that preceded the 2024 carry unwind, and neither the gauge nor the fast-cut watches the carry trade directly. Third, the narrowness of Thursday's rally — megacaps up 1.81% and momentum up 2.10% against an equal-weighted index that lagged badly — is the concentration the model holds twice, through the growth cycle and through the mega-cap sleeve. Fourth, oil: the October contract fell back to $99.53 while the source series rolled early to November, and the inflation composite is holding Thursday's settle rather than a rolled figure. Goldman Sachs, on the desk's own channel, published twice into the weekend: "The Opportunities for Investors amid Higher-for-Longer Interest Rates" (18 September) and "Can Stocks Rally With a Hawkish Fed?" (19 September). Neither clip carries captions, so neither is summarised here beyond its title and date; the substance cited above is the firm's published revision to its rate path. The titles are themselves the week's question, and the gauge answers it mechanically rather than directionally: equities can rally against a tightening central bank for as long as credit stays calm and the trend holds, which is precisely what the fast-cut measures. Both legs are clear this week, and credit is the closer of the two. Mark Meldrum's weekly review (19 September) supplies the sharpest challenge to the decision itself, on two grounds. The first is internal: the Summary of Economic Projections holds unemployment at 4.1%, shows real growth easing rather than rising, has inflation drifting back to target, and still implies roughly 90 basis points of easing later — a set of paths he argues cannot describe the same economy, and which sits oddly beside a decision to tighten now. The second is what the press conference omitted. Asked why the long end keeps rising, Chairman Warsh offered economic strength, competition for capital and geopolitics, and did not mention the fiscal position: interest expense of about $1.4 trillion this year, above 4% of GDP and close to a quarter of federal receipts, rising further after this week's increase. Two of the three explanations also point the wrong way — first-half growth ran at 1.8%, below the committee's own long-run range, and a riskier world usually bids for Treasuries rather than away from them. The model does not price the fiscal path, and no cycle in the gauge would capture a term-premium repricing driven by supply; that gap is stated here rather than solved. His market read corroborates the tape: the thirty-year set a cycle high at 5.37% on 10 September and the twenty-year at 5.40% on 15 September, while the long-duration equity proxies fell with bonds (utilities −3.0%, real estate −2.7%) and the dollar rose on the rate differential.
And the prediction markets, read against our cycles: the crowd prices the Fed hawkish, which confirms our tight monetary read; puts recession at 6% (benign risk), consistent with our conviction (-0.00); puts Hormuz traffic back to normal by year-end at 14% — the supply shock is the base case, a flag against our energy-led inflation read. Net, Polymarket's macro pricing this week and our six cycles corroborate the model's cautious stance.
Prediction markets this week
Monetary · Financial conditions · Dollar — where the crowd is pricing the macro.
Target allocation
| Sleeve | Macro Tide Core | Macro Tide Spring | Macro Tide Harbor |
|---|---|---|---|
| US equity, total | 45% | 45% | 41% |
| SPY (S&P 500) | 31.8% | 31.8% | 28.6% |
| MAGS (mega-cap growth) | 13.6% | 13.6% | 12.3% |
| Int'l developed (VEA) | 11% | 11% | 10% |
| Emerging mkts (EEM) | 12% | 12% | 10% |
| Diversified commodities (PDBC) | 14% | 14% | 12% |
| Gold (GLD) | 13% | 13% | 11% |
| Bitcoin (IBIT) | 2% | 2% | 2% |
| Managed futures (DBMF) | 0% | 0% | 0% |
| Long bonds (TLT) | 3% | 3% | 3% |
| Cash (USFR) | 0% | 0% | 10% |
| Pair overlay — long QQQ | — | +15% | — |
| Pair overlay — short IWM | — | −15% | — |
| Gross exposure | 100% | 130% | 100% |
Three products, one engine. Macro Tide Core — the engine, long-only, for retirement accounts · no withdrawals · drawdown budget 12%. Macro Tide Spring — the engine plus a 15/15 market-neutral pair, in a taxable margin account · gross 130%, net 100%; its sleeve weights are the Macro Tide Core's, the pair sits on top. Macro Tide Harbor — the engine with two years of spending held in T-bills, for clients drawing income · hard 10% cash floor at 5%/yr; 4% / 6% are dials: floor = max(10%, 2 × rate); today the floor binds (the Macro Tide Core holds 0% cash, this book holds 10%). All three: risk cap 1.00, ladder 1.00 / 0.90 / 0.80 / 0.25 / 0.10 on conviction, fast-cut to 35% risk, momentum tilt k .50, Gundlach duration gate, MTUM on while conviction > 0. Adopted 2026-09-13 on fifteen pre-registered bars (changelog §27); the books run from the 8/31 close on these rules (§28).
Pair overlay — Macro Tide Spring only (gross 130%, net 100%): long QQQ 15% / short IWM 15% (standing pair — mega-cap over small-cap, positive in every regime state). Sized as a fixed 15/15 on the book, financed at 60bp on the long and 40bp borrow on the short (pair spec §1); the Macro Tide Core and Macro Tide Harbor carry no pair.
US-equity legs: SPY 70 / MAGS 30 (momentum OFF — conviction -0.00 ≤ 0) The US-equity sleeve is a deliberate, sized bet on the AI capex cycle (MAGS, and momentum when conviction is positive) — an exposure the conviction gate and the fast-cut exist to contain, not a claim that the cycle's returns on capital will hold.
Relative-momentum tilt (allocator 1.16): each non-US risk sleeve × (1 + 0.50·z), z = its 12-month return vs SPY against its own 120-month history, clipped ±2; the risk sleeves are rescaled so the risk budget and cash are unchanged. Read at the 2026-10-31 close: Int'l z +0.9 → ×1.47 · EM z +1.1 → ×1.53 · Commodities z +1.6 → ×1.79 · Gold z -0.5 → ×0.73 · BTC z -0.7 → ×0.67 · Mgd fut z +0.7 → ×1.35 · Long bonds z -0.6 → ×0.71.
Duration gate (Gundlach, allocator 1.18): 10-yr 5.24% vs fair value 4.82% ((Bund + 7-yr nominal GDP growth) ÷ 2; Bund 3.50 (2026-10-09), GDP to 2026-04) → gap +0.42pp → TLT at 34% of its regime weight (scales in from +0.25pp, full at +0.75pp). Own duration only when the bond is cheap to his model; at a fair-value gap long bonds beat cash 40% of the time, above +0.25pp 78%.
Commodity gate (allocator 1.20, changelog §33): the commodity sleeve moves to cash when global liquidity momentum or the monetary read sits one standard deviation below its own 60-month norm. This week: liquidity z +0.07, monetary z -0.43 → gate OFF. Adopted on ten pre-registered bars; its work is in tightening cycles (gated months lost 2.8%/mo in 2007–16).
The $100,000 Macro Tide Core ledger
the engine, long-only, for retirement accounts · no withdrawals · drawdown budget 12%. A share-by-share ledger started at the 8/31/2026 closes. It trades only when a sleeve drifts more than 3 points off target, or once a month, and every trade is dated in the rebalance log. A hypothetical track record, not advice.
| Holding | Ticker | Shares | Price | Value | Weight |
|---|---|---|---|---|---|
| US equity | SPY | 42.43 | $778.57 | $33,037 | 32.6% |
| US equity | MAGS | 191.93 | $73.77 | $14,159 | 14.0% |
| Int'l developed | VEA | 164.84 | $70.35 | $11,596 | 11.4% |
| Emerging mkts | EEM | 180.27 | $66.80 | $12,042 | 11.9% |
| Commodities | PDBC | 718.22 | $19.66 | $14,120 | 13.9% |
| Gold | GLD | 34.03 | $384.58 | $13,086 | 12.9% |
| Bitcoin | IBIT | 55.53 | $46.55 | $2,585 | 2.6% |
| Long bonds | TLT | 9.49 | $77.98 | $740 | 0.7% |
The $100,000 Macro Tide Spring ledger
the engine plus a 15/15 market-neutral pair, in a taxable margin account · gross 130%, net 100%. A share-by-share ledger started at the 8/31/2026 closes. It trades only when a sleeve drifts more than 3 points off target, or once a month, and every trade is dated in the rebalance log. A hypothetical track record, not advice.
| Holding | Ticker | Shares | Price | Value | Weight |
|---|---|---|---|---|---|
| US equity | SPY | 43.05 | $778.57 | $33,518 | 32.6% |
| US equity | MAGS | 194.72 | $73.77 | $14,365 | 14.0% |
| Int'l developed | VEA | 167.24 | $70.35 | $11,765 | 11.4% |
| Emerging mkts | EEM | 182.90 | $66.80 | $12,217 | 11.9% |
| Commodities | PDBC | 728.67 | $19.66 | $14,326 | 13.9% |
| Gold | GLD | 34.52 | $384.58 | $13,277 | 12.9% |
| Bitcoin | IBIT | 56.34 | $46.55 | $2,622 | 2.5% |
| Long bonds | TLT | 9.63 | $77.98 | $751 | 0.7% |
| Pair overlay | QQQ | 20.53 | $751.27 | $15,426 | 15.0% |
| Pair overlay | IWM | -55.30 | $278.94 | $-15,426 | -15.0% |
The $100,000 Macro Tide Harbor ledger
the engine with two years of spending held in T-bills, for clients drawing income · hard 10% cash floor at 5%/yr; 4% / 6% are dials: floor = max(10%, 2 × rate). A share-by-share ledger started at the 8/31/2026 closes. It trades only when a sleeve drifts more than 3 points off target, or once a month, and every trade is dated in the rebalance log. A hypothetical track record, not advice.
| Holding | Ticker | Shares | Price | Value | Weight |
|---|---|---|---|---|---|
| US equity | SPY | 38.13 | $778.57 | $29,687 | 29.3% |
| US equity | MAGS | 172.47 | $73.77 | $12,723 | 12.6% |
| Int'l developed | VEA | 148.88 | $70.35 | $10,473 | 10.3% |
| Emerging mkts | EEM | 162.69 | $66.80 | $10,868 | 10.7% |
| Commodities | PDBC | 639.78 | $19.66 | $12,578 | 12.4% |
| Gold | GLD | 30.63 | $384.58 | $11,779 | 11.6% |
| Bitcoin | IBIT | 49.56 | $46.55 | $2,307 | 2.3% |
| Long bonds | TLT | 8.56 | $77.98 | $668 | 0.7% |
| Cash | USFR | 200.62 | $50.44 | $10,119 | 10.0% |
The books vs SPY
Inception 8/31/2026 · week 6 · $100,000 each · products reconstructed to the 8/31 close on the rules adopted 9/13 · marked to the 10/9/2026 close.
Macro Tide Core, Macro Tide Spring and Macro Tide Harbor are $100,000 share ledgers run from the 8/31 close on the rules adopted 9/13 — their history to that date is a reconstruction from the reads published each week, at Schwab closes, with the same drift rule (changelog §28). SPY is a 100% S&P-500 benchmark on the standard weekly basis — the 8/28 close (769.35) → latest close — matching the widely-quoted +0.11% weekly move. Max drawdown is the realised peak-to-trough of the weekly marks — 4 weeks so far, so it is the live path to date and not the 2007–26 backtested figure; those are in Four ways to hold the same risk. Sharpe and beta are withheld until 52 weekly marks (week 4 of 52): annualised from fewer, they would be a confident number about nothing. Hypothetical, not advice. The Balanced book was closed at the 9/11/2026 mark at $100,155 (+0.16% since inception) when the profile was retired (changelog §12).
| Portfolio | Inception | Value | Since inception | This week | vs SPY | Max DD |
|---|---|---|---|---|---|---|
| Macro Tide Core | 8/31/2026 | $101,365 | +1.37% | -0.71% | -0.39pp | -0.71% |
| Macro Tide Spring | 8/31/2026 | $102,825 | +2.82% | -0.19% | +1.07pp | -0.19% |
| Macro Tide Harbor | 8/31/2026 | $101,202 | +1.20% | -0.63% | -0.55pp | -0.63% |
| SPY (benchmark) | 8/28/2026 | $101,198 | +1.20% | +0.66% | benchmark | -0.34% |
Four ways to hold the same risk — 2007–26
The same sleeves, four ways: buy SPY and hold; the static four-sleeve basket from Part 2 (SPY, QQQ, gold, managed futures, fixed weights); inverse-volatility over those four (quarterly, trailing 12-month vol); and the nine-sleeve regime model as run today (six-cycle conviction, fast-cut, momentum tilt, net of 0bp one-way). Sharpe is excess over T-bills. The eras are a regime split, not out-of-sample validation — every rule was chosen knowing the full history.
| Strategy | 2007–26 CAGR | Sharpe | MaxDD | Era A 2007–16 CAGR | Sharpe | MaxDD | Era B 2016–26 CAGR | Sharpe | MaxDD |
|---|---|---|---|---|---|---|---|---|---|
| SPY, buy and hold | 10.9% | 0.65 | -51% | 6.6% | 0.46 | -51% | 15.4% | 0.85 | -24% |
| Static 4-sleeve basket (Part 2) | 10.2% | 0.88 | -23% | 6.3% | 0.61 | -23% | 14.3% | 1.14 | -13% |
| Inverse-vol, 4 sleeves (qtrly) | 9.3% | 0.87 | -21% | 5.2% | 0.55 | -21% | 13.5% | 1.19 | -11% |
| Macro Tide Core (net 0bp) | 13.8% | 1.36 | -8% | 9.3% | 1.03 | -8% | 18.5% | 1.68 | -7% |
| Macro Tide Spring, gross 130% (net 0bp) | 14.3% | 1.39 | -8% | 10.0% | 1.08 | -8% | 18.8% | 1.71 | -6% |
| Macro Tide Harbor, floor 10% (net 0bp) | 13.0% | 1.37 | -8% | 8.7% | 1.04 | -8% | 17.4% | 1.68 | -7% |
Backtest 2026-10-11: neutral risk step 0.55, momentum tilt mom. Inverse-vol is the honest passive alternative: it keeps most of the static basket's return with less drawdown, and the regime model's edge over it is the drawdown, not the return.
Spending from the books — the Macro Tide Harbor and the others under withdrawals
The Macro Tide Harbor is built for a 5%/yr withdrawal: the same engine with a hard 10% cash floor (2 years of spending) that tops the cash sleeve up, never adds to it. The glide never spends the cushion: when the model returns idle cash to the risk sleeves after a bottoming signal, it deploys only what sits above the floor, so the 2 years of spending are intact in every month of both test samples. The other lines are shown under the same withdrawals with no floor of their own. Below: withdraw the dial's rate of starting capital every year (monthly, not inflation-indexed) from 2007 to today, and over every rolling 10-year window. Start 100. The Conservative and Balanced profiles were retired (changelog §12).
| Dial | Line | End 2007–26 | Worst drawdown of wealth | Rolling 10-yr end, median | Rolling 10-yr end, worst | Rolling worst drawdown |
|---|---|---|---|---|---|---|
| 4% | Macro Tide Harbor (floor 10%) | 736 | -9% | 218 | 171 | -9% |
| Macro Tide Core, no floor | 873 | -9% | 232 | 178 | -10% | |
| Macro Tide Spring, no floor | 966 | -9% | 252 | 198 | -10% | |
| SPY, buy and hold | 402 | -54% | 265 | 119 | -54% | |
| 5% | Macro Tide Harbor (floor 10%) | 641 | -9% | 198 | 154 | -10% |
| Macro Tide Core, no floor | 768 | -9% | 211 | 161 | -10% | |
| Macro Tide Spring, no floor | 855 | -9% | 231 | 181 | -10% | |
| SPY, buy and hold | 309 | -55% | 246 | 100 | -55% | |
| 6% | Macro Tide Harbor (floor 10%) | 546 | -10% | 179 | 138 | -12% |
| Macro Tide Core, no floor | 662 | -10% | 192 | 144 | -12% | |
| Macro Tide Spring, no floor | 744 | -10% | 210 | 163 | -11% | |
| SPY, buy and hold | 216 | -56% | 227 | 81 | -56% |
Nominal withdrawals of starting capital; indexing them to inflation makes every line harder, the 6% dial most. The floor costs a little return in high-conviction months (it binds only when the model wants less than 2 years of spending in cash) and buys liquidity, not drawdown protection: the client never sells the engine to fund a withdrawal. Backtest 2026-10-11.
Commentary
Not investment advice. Hypothetical framework output computed on public data (FRED + Yahoo Finance); past performance does not predict future results. This is a research and education note, not a recommendation.
© 2026 Nemdhari LLC. All rights reserved. The Macro Tide framework — its six-cycle regime gauge, conviction score, allocation rules, indicators, code and model portfolios — is the proprietary intellectual property of Nemdhari LLC. Research and education only; not investment advice, not an offer or solicitation, and not a recommendation for any person. Hypothetical and model results do not reflect actual client accounts; past performance does not predict future results.
Also on Substack · bambi9.substack.com