Trend Following (4/4): The Poor Man’s Trend Program
The first three parts of this series built a trend-following program and took it apart: sixty-two futures markets replicated, distilled to one contract per sector, then measured for what trend actually adds to a risk-premia core. All of it sized to a volatility target, indifferent to the account behind it. This closing part asks the question that indifference skips — what can a private investor actually hold? The answer is more generous than the research let on. Sized honestly by risk, the full ten-sector program reaches a floor near seven hundred thousand dollars — an affluent private account, not a fund, and within reach of someone late in their working life or newly retired; and below that line the move is not a watered-down subset but the core itself, tilted asset by asset by its own trend — cheap insurance against the one regime it cannot survive alone.
Ten contracts, and what they cost to hold
Part 2 already did the hard compression. The full universe collapsed to one representative contract per sector — ten contracts that reproduce the program’s risk-adjusted return, because its risk lives in roughly ten independent directions and one clean instrument spans each. That ten-contract book is the implementable program. The question is whether even ten contracts can be held at retail, and the answer is yes. The constraint is granularity, not access — and granularity sets the size of the account, not a wall in front of it.
The sizing test is concrete. Each contract carries a dollar risk equal to its notional times its volatility, and a sector can be traded cleanly only when the account’s risk budget for that sector covers several whole contracts — otherwise rounding to an integer swamps the signal. The representative volatility is the annualized standard deviation of each market’s daily returns over its price history — about sixteen percent for equities, five for ten-year notes, twenty-five for copper, fifty-odd for bitcoin. The per-sector budget is not free to choose: ten roughly independent sectors sharing an eight-percent target each carry 8% ÷ √10 ≈ 2.5% of volatility. Require room for at least two whole contracts — enough to size a sleeve in steps rather than switch it on and off — and the binding market is the one with the largest risk per contract that has no smaller version.
That reorders intuition. Bonds, despite their size, are among the easiest to place: their volatility is low, and a ten-year micro-yield future trades the rate directly for a few thousand dollars of risk. The real constraints are meats and softs — no micro contracts, moderate volatility, so lean hogs and sugar set the floor. On these terms the full ten-sector program reaches a floor near seven hundred thousand in capital. That figure is idealized: it assumes the sectors are perfectly independent, so each needs only 8% ÷ √10 of volatility, and it accepts coarse two-contract sizing. Any real correlation between sectors lowers their effective number and pushes the account up, as does finer sizing — in practice the number lands nearer one to one-and-a-half million. But the floor is the point: this is an affluent private account, not a fund, within reach of someone late in their working life or newly retired, who is exactly the investor that most needs what trend does.
Breadth is the Sharpe
The reflex is to keep only the few sectors where trend looks strongest and call it a cheap version of the program. It is not. The program’s Sharpe is a breadth phenomenon — no single sector’s sleeve is impressive on its own, and the return appears only as ten weakly-related sectors accumulate. Worse, the highest single-sector Sharpe sits in equities, a market the core already holds outright, so concentrating in the high-Sharpe sectors loads up on precisely the exposure that is redundant with the core. Breadth cannot be faked with a handful of favourites, and the favourites are the wrong ones to pick.
So with around a million in capital you can comfortably run the whole program — less at its idealized floor; below that you cannot fake it by hand-picking the sectors you can afford. The honest move under that line is not a subset of the program but a different object entirely — the long-only core, tilted by its own trend — and that is what the rest of this part builds.
The rule: the core, tilted by trend
The construction reuses the passive core’s machinery entirely, so it is worth stating exactly. For each of the three assets, take two numbers straight from RP3: its inverse-volatility weight, and the single portfolio scalar that targets eight percent volatility — the same covariance-free sizing this series has used throughout. Trend enters only as a tilt on top. Writing the trend signal as a number between minus one and plus one, each asset is held at its passive weight × the vol scalar × (0.5 + 0.5 × trend). The core supplies a baseline of one half; the trend tilt adds or subtracts up to one half on top. An asset trending up is carried at its full risk-premia weight, one with no trend at half, one trending down is cut to zero.
Two properties follow, and they are the whole point. Nothing is ever shorted — the multiplier never goes below zero. And because the trend rides the core’s own sizing rather than chasing a volatility target of its own, nothing is ever levered: when conviction is weak the book simply holds less. Equivalently, the combined strategy is one half the passive core plus one half a trend-tilted copy of it, the two sharing a single set of weights and a single vol target. That is the strategy plotted as “RP3 + trend” throughout what follows; the passive core is RP3 with every trend tilt set to its neutral half.
One more rule makes the figures honest about cost. Because the tilt usually holds part of the book in cash, that cash is credited the Fed Funds rate; on the rare days the vol scalar pushes the book above one times capital, the borrowed slice is charged Fed Funds plus one and a half percent, the standard Interactive Brokers margin spread. Every return that follows is net of this financing. It cuts both ways: the de-risked book earns interest on its idle cash — real money, but in the high-rate decades a large one — while a vol-targeted passive core, which quietly levers toward one and a half times in calm markets, pays to do so. That cash carry flatters the deep-history Sharpe: the 1.50 in the figure above falls to 1.12 once it is stripped out, nearly all of the difference earned in the 1980s when Fed Funds ran near ten percent and the de-risked book sat half in cash. The scaled version later, which holds far less cash, is the number to trust.
It is regime insurance
The combined book does not beat the core evenly through time — it beats it exactly where the core is exposed. A risk-premia core leans on stocks and bonds offsetting each other; when their correlation turns positive that internal hedge is gone, and the strategy’s value-add concentrates almost entirely in those positive-correlation regimes. The episode that matters is the stagflation corner — positive correlation with falling equities, the nineteen-seventies and 2022 — where the core has nothing but gold and the overlay quietly steps out of the falling assets. In the long disinflationary bull, by contrast, the overlay is a gentle drag: it de-risks and gives up some of the ride. That is not a flaw. It is what insurance does.
Stock-bond correlation (top), trailing equity return (middle), and the strategy’s value-add over the core (bottom). The bottom panel is the 3-year rolling difference in annual return, scaled strategy minus passive core, both at ~8% volatility, so it reflects regime timing rather than a vol gap. Shaded: the stagflation corner. The value-add runs strongly positive there and only mildly negative through the long disinflation bull — the signature of a hedge.
And there is a reason the case rests on the deep history. The stagflation corner is barely two percent of days after 1995 — essentially just 2022. The regime that justifies trending your own assets is almost entirely a feature of the decades before the modern sample, which is exactly why the fifty-seven-year window matters: it is the only window in which the insurance is seen to pay.
The modern reality
Look only at the period that uses real ETF tickers throughout, from 2008. Here the honest verdict is plain. The combined strategy gives up a little Sharpe to simply holding the core — 1.08 against 1.03 — and in exchange holds a shallower worst drawdown, ten percent against sixteen. In an era with almost no stagflation and a relentless bond-and-equity bull, the tilt has mostly sat idle, costing a small premium and paying little. That is not an argument against it. It is what every insurance policy looks like in the years before the event it is written against.
Using the full risk budget
One honest weakness remains. Because the tilt averages the book down, the combined strategy runs near five and a half percent volatility — it leaves a third of its risk budget unused, and gives up compound return to do so. The fix is not to vol-target on a short window, which would lever straight back into the weak-signal trap. It is to scale once, slowly. Take the realized volatility of the combined book over a trailing ten years and lift the whole thing to eight percent. A decade of volatility cannot react to a quiet quarter, so the scalar is strategic, not tactical: it corrects the chronic under-exposure and then holds.
Across the full span the scalar sits near one and a half times and barely moves. The ten-year window only populates around 1980, so for the burn-in before it — the 1970s, the very decade the strategy is built for — the scalar is simply held at that same long-run one and a half. With the stagflation decade restored, the scaled book runs at Sharpe 1.23 against the core’s 0.95 since 1970, compounds at 10.1% against 8.1%, and still draws down less than half as deep — fifteen percent against thirty-two — all net of financing. The figure also lays the plumbing bare: the position sizing that stays mostly below one times capital, and the cumulative interest earned on idle cash against the smaller sum paid to borrow. This is the version to actually run — the insurance, scaled to pull its weight.
What this actually is
The full-breadth trend program is not the institutional object it looks like. Sized by risk rather than notional — and using a yield future for the bonds — all ten sectors clear at an idealized floor near seven hundred thousand in capital, and in practice somewhere around one to one-and-a-half million — an affluent private account, not a fund, within reach of someone late in their working life, precisely the investor who most needs what trend does. Above that line, run the whole program; you are priced out only of the institutional contract list, not of breadth itself.
Below it, the move is not to hand-pick the few sectors you can afford. Breadth cannot be faked, and the high-Sharpe sectors are the ones the core already owns. The move is the strategy built here: the long-only risk-premia core, tilted asset by asset by its own trend, scaled to use its full budget. Three ETFs, net of financing, ahead of the passive core over every long window — not by predicting anything, but by stepping out of what is falling and holding what is rising.
And this is the part to be clear about. It is not another tactical asset-allocation overlay bolted onto a model portfolio. It is the top of a full systematic trend program — the same sixty-two-market engine of the first three parts — broken all the way down and recombined with a long-only portfolio until what remains fits in a brokerage account. The pedigree runs top-down, not bottom-up.
We do not know whether the next decade looks like the 1970s. That is precisely why you hold this. The tilt costs little through the calm regimes and earns its keep in the one the core cannot survive. Carry it for the decade you cannot forecast.
Ah ok ill go back and reread that one! Appreciate the reply. Thank u
Nice post; thx for sharing. What if you ran the enhanced core (meaning the inverse vol with trend scaled up) + the separate trend program? So like combining both: the dynamic trend-informed beta + the replicated trend overlay.