Managed futures are the textbook answer to "what saves my portfolio when stocks and bonds fall together". I hold them myself. But they come with real friction: 1-2% fee layers, opaque holdings, and a return stream that spends years looking broken before it pays off. So when the Beyond Passive newsletter published a series called "the poor man's trend program", asking how close you can get with three plain ETFs and one trend rule, I had to check the numbers myself. Full credit where it's due: the idea is theirs, not mine. I rebuilt it on our engine and data to see if it survives.
The recipe
Start with a three-asset risk-parity core: US stocks (SPY), long Treasuries (TLT), and gold (GLD), each weighted by inverse volatility so the shaky asset gets less and the calm one gets more. Then tilt each asset by its own trend, once a month:
- Uptrend: hold the full risk-parity weight
- Neutral: hold half
- Downtrend: drop to zero, park the slice in T-bills
That's the whole strategy. No leverage, no futures account. It never holds more than about half the book in equities.
What my replication found
One catch: the article deliberately withholds the exact trend formula. A strategy that only works with one secret signal is a curve-fit, so I tested four common trend definitions instead, from a simple "is the 12-month return positive" to a multi-speed z-score. On proxy data back to 1972 (gold only floats freely from late 1971), here's what came out:
| Version (1972+) | Return/yr | Sharpe | Worst drop |
|---|---|---|---|
| Static risk-parity core | 5.6% | 0.95 | -37% |
| Core + one 10-month SMA gate on the whole book | 5.6% | 1.22 | -27% |
| Per-asset tilt, binary 12-month momentum | 7.8% | 1.51 | -12% |
| Per-asset tilt, risk-adjusted 12-month momentum | 7.5% | 1.56 | -11% |
| Per-asset tilt, 10-month SMA band | 7.1% | 1.50 | -13% |
| Per-asset tilt, multi-speed z-score | 6.3% | 1.59 | -11% |
Two things stand out. My static core lands on a 0.95 Sharpe, matching the article's own figure exactly, which gave me confidence the comparison is apples to apples. And every one of the four trend definitions cut the worst drawdown from -37% to roughly -12%. Any sane trend measure captures the edge, so the withheld formula can stay withheld...
Why per-asset beats one big switch
The lazy version of trend protection is a single gate: when the S&P drops below its 10-month average, sell everything. That helps (the -37% becomes -27%), but it has a blind spot. The gate watches stocks, so it sleeps through a bear market in bonds. Anyone holding TLT through 2022 knows what that feels like, and the 1970s version ran for a decade. Tilting each asset by its own trend means the bond sleeve can go to cash while stocks and gold keep running. That independence is where the extra Sharpe comes from.
The honest caveat
Would I expect the full 1.5 Sharpe going forward? No. Most of the outperformance sits in the 1970s to 1990s: stagflation, a long bond bear, trending gold. Since 2005 the tilt hasn't beaten the static core on Sharpe. It still cuts the drawdowns, but it pays for that with return, like any insurance. If you expect the next 20 years to look like the last 20 (calm, stock-led, quick V-shaped recoveries), the static core is fine and cheaper to run. The tilt earns its keep in the regimes everyone hopes never come back. I think that's exactly why it deserves a place in the catalog, and also why I won't sell it to you as a return booster.
Try it
The strategy is live as RP3 Trend-Tilted Core, in two variants (risk-adjusted momentum and SMA-band). The live backtest runs from 1973 and currently shows 8.4% a year at a 1.17 Sharpe with a -12.4% worst drawdown, checked monthly like everything else on the platform. The original write-up is at Beyond Passive and it's worth your time.