When the Macro Tide Actually Moves the Boats

Where we left offTwo parts, one blind spot

Part 1's lesson was that you can manage risk but not manufacture return: a 10-month trend switch cut drawdown roughly in half for a small give-up in return, and hit a ceiling around a 0.74 Sharpe. Part 2's lesson was bigger — diversification is the free lunch. An all-weather basket (stocks, gold, managed futures) matched SPY's return with half the drawdown and a Sharpe near 1.0, no market-timing required. That basket is the engine.

But both Parts had the same blind spot: they only ever looked at price. Part 1's moving average is a lagging line; Part 2 doesn't look at the outside world at all. So Part 3 asked the obvious question a professional macro investor would ask: is there information outside of price — in liquidity, in yields, in credit — that leads the market, and can we use it?

The one ideaMarkets move on the discount rate

Here's the mental model the whole thing rests on, and it's worth internalizing even if you never build a model:

That last part — liquidity — is the big one, and it has a special property. Liquidity isn't cash; it's credit, most of it created against collateral, and it feeds on itself (rising asset prices become collateral for more borrowing, which buys more assets). Crucially, that credit shows up in the financial plumbing before it shows up as buying pressure. The pioneer of this view, Michael Howell, finds that global liquidity leads risk assets by about 13 weeks — a full quarter of warning, if you can measure the tide.

We measured it, with the public proxy for the cash available to markets: Net Liquidity = the Fed's balance sheet − the Treasury's cash account − the reverse-repo drain. And the 13-week lead is real: liquidity momentum today correlates about +0.35 with the S&P's return a quarter out. Real — but weak. Hold that number; it turns out to explain everything that follows.

Building the dialTwo axes, then four cycles on top

Turning that into a usable read meant building a compact, transparent gauge from data anyone can download from the St. Louis Fed. It starts with two axes:

• A growth axis — is the economy accelerating or decelerating? Here we made a deliberate choice that matters in 2026: we refused to let the gauge be blind to the AI-capex boom. The old-economy proxies read the economy as soft — but ~$1 trillion a year of AI investment is the real growth engine, so we fold in equipment orders and power-grid construction to capture it.
• An inflation axis — accelerating or decelerating?

Together these give the classic four-regime map — Goldilocks, Reflation, Slowdown, Stagflation. But a regime is a snapshot, and a snapshot doesn't tell you how hard to bet. To size a position you need to know which way the deeper currents are running, so the gauge tracks four more cycles that ride on top of the map — six in all, the same multi-cycle architecture the sharpest macro shops (Darius Dale's 42 Macro among them) run:

• Liquidity — the 13-week lead, risk-on or risk-off. The big one.
• Monetary — where the market thinks the Fed is going, read straight from prices (the 2-year yield, SOFR, the real policy rate), cross-checked against live prediction-market odds. This cycle does double duty: it doesn't just weigh on conviction, it gates duration — when policy reads tight, we refuse to hold long bonds.
• Positioning — the contrarian cycle. When everyone is already leaned the same way — a sleepy VIX, price stretched far above trend — that crowding is the risk, and it quietly trims how hard we press.
• Financial conditions & the dollar — real yields, and whether the dollar is weakening. A falling dollar is the debasement signal the whole thesis points to; it tilts the mix toward gold (a real asset, not a safe-haven). That tilt is conditioned on liquidity, following Howell: gold is held through the cycle, but its liquidity-sensitive cousins — emerging markets and bitcoin — are only pressed when the tide is rising, and pulled back into gold when it ebbs.

Sitting across all of it is a stress veto — if the bond market's fear gauge blows out, everything goes defensive regardless of what the cycles say. Getting this to behave took real iteration, and much of the work lives in those details.

The moment of truthA humbling first result

Then we backtested it, 2007–2026, across three investor profiles. The first result was deflating: the regime-tilted portfolio's risk-adjusted return was identical to just holding the same assets at fixed weights. We ran the fair test — regime-timing versus the exact same sleeves held static — and timing was a wash, and worse, inconsistent: it hurt through the 2008 era and helped modestly since. Mechanical macro tilting, it turns out, is a way to feel sophisticated while adding nothing.

That should have been the end of it — and it lines up perfectly with the academic record, where decades of work (Welch & Goyal, most famously) shows macro predictors that dazzle in-sample usually fail out-of-sample. Even Druckenmiller says the signals he traded for decades are now "20% as effective," loved to death by everyone using them.

Doing it the way it's doneAsymmetry, and the instant cut

But a mechanical tilt is not how a real macro investor operates — and that's the insight that rescued the project. Stan Druckenmiller's method, articulated across his interviews, is built on asymmetry:

You bet big only on the rare one-way bets where everything aligns, and you change your mind instantly — cut the moment the thesis breaks. So we rebuilt the overlay his way. Instead of always tilting a little, we sized by conviction. The result was electric — and terrifying. In the modern era it was among the best we'd tested; but through 2008, without a cut, it cratered around −30%. Asymmetry is a loaded gun: big and wrong is a catastrophe.

The fix was the other half of his method — cut instantly. The moment liquidity turns down, or credit spreads widen, or SPY breaks its 10-month trend, de-risk immediately, without waiting for the slow monthly regime to flip. That single discipline changed everything:

With a full-history credit trigger — the fast-cut fires on a genuine 2008/2020 credit blow-out, not just the trend break — the cut roughly halves the drawdown again, to ~−9% full-cycle, and now edges the fixed-weight baseline on Sharpe too. Folding in the two market-priced cycles — monetary and positioning — adds a little more, for free. But keep the honest anchor: the static Part-2 basket still carries the best full-cycle Sharpe (0.88). The overlay's win is drawdown, not return.

The sharpest refinementDon't hold the asset designed to lose

There is one more refinement, and it is the clearest illustration of what the whole machine is for. Look at what the monetary cycle was telling us: policy tight, higher-for-longer, long yields at risk of rising. And look at what a conventional "defensive" playbook says to hold in that environment: long-dated Treasuries — the one asset that loses precisely when long yields rise. That is a contradiction, and the framework should not tolerate it.

So we gated duration on the monetary cycle. When the read is tightening, the long-bond sleeve is rotated out — into short-dated T-bills (which earn the high short rate, income without the duration risk) and a little equity — and it rotates back the week the monetary signal flips toward easing. It is the same asymmetric discipline applied to a single sleeve: don't own the thing the cycle says will fall; step back in the moment it says the coast is clear.

This is where the refinements compound. In the modern liquidity-and-rates regime (2016–2026), the finished strategy reaches a Sharpe of ~1.05–1.11 at a worst drawdown of ~−7% — running roughly even with the static Part-2 basket on Sharpe (its 1.13) at about half its drawdown (its −13%). Dodging the 2022 long-bond crash is a large part of that. The honest asterisk stays: full-cycle, the basket's Sharpe (0.88) is still higher than ours (0.79); the overlay's edge is the drawdown, not the return. Regime-conditional — but the regime it is conditioned on is the one we are living in.

And it is worth naming the strongest objection, from a direction I respect. Apollo's Torsten Sløk, writing the same week, argues long rates are set to fall over the coming months whatever the AI outcome — if AI delivers, it is deflationary; if it disappoints, a flight into Treasuries does the same. If he is right, gating out of long bonds is a mistake, and that is the honest bull case for the very asset the rule just sold. Two things let me hold the position with eyes open. First, the gate is not a permanent short: it steps back into long bonds the week its signal turns, so a real decline in yields pulls us back in with about a week's lag. Second, the exact variable that would prove Sløk right — a falling term premium at the long end — is already on the dashboard, and today it is rising, not falling. The gate is a bet made in full view of the case against it, and built to change its mind quickly.

A live stress testA Fed chair and a legend, the same week

A model is only as good as its reads survive contact with the real world, so consider the week this was finalized. Two of the most consequential voices in macro spoke — independently, from opposite institutions — and both corroborated exactly what the machine was already doing.

On August 28, 2026, the new Fed chair, Kevin Warsh, used his Jackson Hole keynote to say the quiet part plainly: inflation is above target and prices are the Fed's predominant focus, growth and AI-driven capital spending are strong, and — pointedly — he is "hard pressed to describe broad financial conditions as restrictive." He rejected forward guidance and left a conditional tightening bias in the air. That is the monetary cycle's −0.74 read in prose, and the clearest possible justification for the duration gate being off.

Four days earlier, Stanley Druckenmiller — whose method this entire strategy is built on — published "Let the Bond Market Speak." His subject was the Treasury's decision to double its long-bond buybacks off-cycle, days after the 30-year yield hit a 19-year high. He called it price management dressed as liquidity support — a stealth dose of quantitative easing run out of the Treasury while inflation sits above target — and warned that long yields want to rise:

Sit with the symmetry. The framework is built on Druckenmiller's own discipline — bet asymmetrically, cut instantly, don't own what's set up to lose — and the monetary-gated duration rule had just, on exactly that logic, rotated the portfolio out of long bonds. Then Druckenmiller himself published a warning about the very asset the rule had exited. We didn't build the gate because of the op-ed; we built it from the monetary cycle, and the op-ed arrived as confirmation. That is what validation is supposed to look like.

There was one honest lesson in it, too. Druckenmiller's "QE run out of the Treasury" is a liquidity action our Fed-centric proxy under-measures, because it is the Treasury removing duration, not the Fed expanding its balance sheet — the "Treasury twist" we'd flagged as a blind spot. But note the direction: stealth easing into above-target inflation is not a risk-on green light. It is fuel for the debasement trade the portfolio already leans into — real assets and gold over suppressed bonds — which is why the gate and the fast-cut, not the liquidity tailwind, stay in charge.

Why it worksStraight out of the mechanics

Why asymmetry? Because the liquidity signal is weak on average (that +0.35) — noise most of the time, genuinely strong only in the rare moments when everything aligns. A steady tilt dilutes the few good signals with the mass of noise and nets to a wash. Asymmetry concentrates your risk into the moments the edge actually exists.

Why the instant cut? Because liquidity-driven markets are reflexive — the same collateral loop that lifts prices runs in reverse on the way down and cascades faster than fundamentals ever could. If you're big and wrong, hoping is how you go broke. The cut is the circuit breaker.

Why does it work best now? Because the whole edge depends on liquidity being the dominant driver — true in today's post-QE world, and not true in the 2008 deleveraging. The tool works in the regime it was built for, and you must never assume that regime is permanent. So the fast-cut stays always on — insurance for the day the tide goes out for good.

Show your work None of these numbers are a black box. Every score is a transparent weighted average of public data — this week's +0.03 conviction and −0.74 monetary reads are walked through line by line, table and all, in The Gauge Math.

The synthesisIt was all one thing

Here is the quiet beauty of where we landed. Look at what the winning strategy actually is:

The series didn't end with three separate ideas. It ended with one — and each piece covers the others' blind spot. Part 2's engine gives you something worth steering; Part 3's conviction tells you when to press; Part 1's trend-break tells you when to stop.

What it's actually forNot alpha — something better

The product here is not "beat the market with macro." Full-cycle, the static Part-2 basket still carries the best Sharpe (0.88); diversification remains the engine, and anyone who claims macro-timing is a reliable source of full-cycle alpha is overstating the case. What this framework does deliver is three things a single static basket cannot:

• Personalization — the same regime read produces three genuinely different portfolios, across nine sleeves. These are risk tiers, not vol targets (there is no vol-scaling engine): realized full-sample vols were about 5% / 6–7% / 7–8%, with max drawdowns near −11% / −9% / −9%. Pick your tier, and the machine sizes everything to it.
• Drawdown control that survives regime changes — ~−9% full-cycle through 2008, 2020 and 2022, versus SPY's −51%. The fast-cut (with a full-history credit trigger) and the duration gate buy that.
• A competitive read in the current regime — in 2016–2026 the Balanced profile posts about a 1.05 Sharpe at a −7% drawdown — roughly even with the diversified basket on Sharpe, at half its drawdown. A favorable trade, not a claim to beat diversification; regime-conditional, not permanent.

That is a "pick your risk level, then sleep through the drawdowns" tool, built on a diversified engine and steered by a deliberately modest macro signal — a more defensible, and more useful, proposition than a claim to alpha.

A hierarchy of ideasEach rung better than the last

The quiet irony of the whole series: I set out in Part 1 to out-time the market, hit a wall, and spent three parts learning that the edge was never a cleverer signal. It was a portfolio you diversify, a conviction you size into, and the discipline to cut when you're wrong — the same three lessons the best macro investors have been preaching for forty years.

CaveatsBecause this is still a backtest

The method is the point. None of this was stumbled upon: we reasoned from the mechanics of liquidity, real yields and the discount rate to a strategy, tested it against its own fair baseline, watched it fail as a mechanical scheme, and made it work only by running it the way the discipline's best practitioner does — and we keep the limitations in plain view. That process — not the final Sharpe — is what is worth keeping.