"Dual Momentum beats Buy & Hold on return and drawdown." You read that sentence in every other strategy book. Gary Antonacci made it famous with Dual Momentum Investing, and Meb Faber's GTAA work shows the same pattern: cross-sectional momentum — buy the strongest assets, rotate monthly, move to cash when the trend rolls over — is supposedly one of the few anomalies that stays robust across decades.
We rebuilt it. Not with marketing numbers, but with the same code that powers our backtests.
The setup
We use the classic GTAA-5 universe — five ETFs covering five asset classes:
| Ticker | Asset class |
|---|---|
| SPY | US equities |
| EFA | International equities |
| VNQ | REITs (real estate) |
| AGG | Bonds |
| GLD | Gold |
The rule is textbook dual momentum:
- Cross-sectional: at each month-end, rank all five by 12-1 momentum (return over the last 12 months, skipping the most recent month — the standard reversal guard).
- Hold Top-N: hold the strongest N assets equally weighted, rotate monthly.
- Cash gate (absolute momentum): if the chosen asset's momentum is ≤ 0, go to cash instead of holding. This is the trend filter that protects against bear markets.
Look-ahead-free: ranking runs on the completed month-end bar, the rotation is captured the following month. No peeking into the future.
Test window: January 2007 to June 2026 — 233 months, roughly 19.5 years, including the 2008 financial crisis, the 2020 COVID crash and the 2022 rate-driven bear market.
The result
| Variant | CAGR | Max drawdown | Sharpe |
|---|---|---|---|
| Buy & Hold S&P 500 (SPY) | 10.9 % | −50.8 % | 0.67 |
| Rotation Top-3 (cash gate) | 8.3 % | −17.7 % | 0.67 |
| Rotation Top-2 (cash gate) | 8.5 % | −20.2 % | 0.64 |
| Buy & Hold equal-weight 5 | 7.4 % | −35.7 % | 0.61 |
Read the first two rows side by side. Buy & Hold S&P 500 returns 10.9 % per year. The sophisticated rotation returns 8.3 %. The plain index wins the return race by 2.6 percentage points a year — over 19 years that is an enormous gap.
That is the uncomfortable truth. The most famous tactical strategy in the world does not, over the testable window, beat the dumbest conceivable approach — "buy the S&P and do nothing" — on return.
Where Dual Momentum actually delivers
Now look at the drawdown column. Buy & Hold S&P 500 had to sit through a 50.8 % loss at the worst point. Rotation Top-3: 17.7 %. That is one third of the pain.
And the Sharpe — return per unit of risk — is exactly identical at 0.67. Achieving the same risk-adjusted return at one third of the maximum drawdown means: Dual Momentum doesn't sell outperformance. It sells a calmer night's sleep.
That is a perfectly legitimate product. For anyone who can't survive a 50 % drawdown psychologically or financially (and most investors bail at exactly the low), the rotation is objectively better — even though the ending balance is lower. It's just that "you give up some return but a lot fewer nerves" is a more honest pitch than "beats the market."
Why doesn't it beat the market?
Three reasons, and all of them belong in an honest assessment:
1. The edge comes from data we can't test. Antonacci and Faber show their strongest numbers over 1970–2010. That era — the stagflation 70s, the lost decade of 2000–2010 — is exactly when US Buy & Hold was periodically dead and a trend filter shone. But the GTAA ETFs only exist from ~2005 (AGG 2003, GLD/VNQ 2004). We simply cannot reproduce the literature window. Our 2007–2026 window is dominated by the largest US equity bull market in history — the single hardest environment for any tactical strategy.
2. One window is an anecdote, not proof. A single period on a single fixed universe says little about robustness. It is evidence that the strategy doesn't outperform in the modern US bull market — not that it is broken in general.
3. We flattered the costs. The Top-3 variant had 139 position switches over 233 months. Every switch costs spread, fees and — in a taxable account — realized gains. We modeled zero transaction costs. With realistic costs the already-thin return gap would widen further.
The link to our 200-day strategy
We ran into the same realization recently with a completely different strategy. Faber's 200-day trend regime — invested above the 200-day line, cash below — shows exactly the same profile on broad index ETFs: roughly Buy & Hold return, huge drawdown reduction, neutral Sharpe.
There seems to be a pattern stronger than any single strategy: on broad, long-term-rising equity indices, trend timing does not buy you extra return. It buys you drawdown protection. Once you internalize that, you stop hunting for the strategy that "beats the market" and start choosing the strategy whose risk profile matches your stomach.
(There is one exception: on crypto, where bear markets trend deep and hard, the 200-day regime does beat Buy & Hold on return too. But that is a different story — and a different asset class.)
FAQ
So is Dual Momentum nonsense? No. It's solid drawdown protection at an identical Sharpe — just not a return miracle. The flaw isn't in the strategy, it's in the expectation that gets sold with it.
Would different parameters have given a better result? On paper, probably — we tested Top-1, Top-2, Top-3 with and without the cash gate, and they all land in the same corridor (8–9 % CAGR, much lower drawdown). But any parameter search for the prettiest backtest is overfitting. The honest answer is the unoptimized one.
Why not test Antonacci's original window? Because the ETFs only exist from ~2005. We could splice index data together, but that's a different data source with its own biases. We'd rather show honestly what real, tradeable ETF prices give you — and state clearly that our window doesn't contain the literature edge.
Does anything beat the S&P 500 long-term on return? In the equity bull market of the last 15 years: very little, reliably. Which is exactly why "beats Buy & Hold" is a red flag whenever a strategy advertises it. Always ask: over what window, in which asset class, after costs?
Does this count as statistically robust? With caveats. 233 months is a decent sample, but 139 position switches and a single universe remain one path. We treat the result as a strong indication, not final proof.
What Backtesting Arena contributes
We don't build strategies because they're famous — we build them only after checking whether they deliver an edge on real prices. Dual Momentum failed that test as a pure return product, so we didn't ship it as a feature; we published the numbers instead.
You can trace exactly that logic in the Arena yourself: every backtest shows you not just the return, but the drawdown and the Sharpe against Buy & Hold. And our risk-adjusted scorecard tells you straight whether a strategy truly outperforms or merely lowers the drawdown. That distinction is the whole point.