How it works

Monthly rotation across 11 global asset classes, based on Meb Faber's Global Tactical Asset Allocation (GTAA, 2007). Every first trading day of the month: rank all assets by momentum, select the top N, filter out those in a downtrend or with negative momentum, and hold cash for the rest.

The 5 variants

AGG 1 through AGG 5 represent the same strategy with different concentration levels. AGG 1 holds 1 asset (100% in the top pick), AGG 5 holds 5 assets (20% each). Lower numbers are more concentrated and volatile; higher numbers are more diversified and smoother.

AGG 3 is the original Faber configuration. AGG 1 is the most aggressive — pure momentum conviction in a single asset class. AGG 5 is the most defensive — nearly half the universe gets capital, reducing single-asset risk at the cost of diluting the momentum signal.

The consolidated universe

The original Faber universe had 13 assets with a heavy bond bias (6 out of 13 were bonds). In a falling rate environment, the top 3 were often 2-3 bond funds — that's a bond portfolio, not diversification.

The consolidated version groups correlated pairs into 10 independent asset classes — US Equity (SPY/QQQ/IWM), Treasuries (IEF/TLT), and Credit (LQD/HYG) each become a single slot — and adds an 11th slot for foreign sovereign bonds (IGOV signal → IGLA UCITS) to cover USD-weak regimes that pure USD-denominated assets cannot express. The ranking universe is balanced across equities, real assets, real estate, USD fixed income, and foreign sovereign.

Best horse rule

For the 3 consolidated groups, the ranking decides "should I be in this asset class?" and the best horse rule decides "which ETF has the most momentum right now?"

In practice, QQQ gets selected over SPY about 86% of the time when US equity is in the top N. TLT dominates IEF 5:1 for treasuries. This adds a few basis points of CAGR by riding the more aggressive instrument within each correlated group.

Dual filter

Each selected asset must pass two filters before receiving capital: price above its 10-month moving average (absolute momentum), and average momentum above zero. Both filters avoid riding declining assets. During 2008, the strategy was fully in cash before Lehman collapsed.

Stress testing

Validated across 78 parameter configurations with a CV of Sharpe = 17%. No single asset can break the strategy (max impact of removing any one asset: 0.08 Sharpe). Smaller universes sometimes perform better. The strategy works at every rebalance day tested (1st-20th of the month), though the first few days are consistently better.

UCITS execution

Trades are executed using Irish-domiciled UCITS funds, mostly ETFs on the London Stock Exchange (IPRE trades on Xetra in EUR; SGLD is an ETC, not an ETF). For most non-US persons this sidesteps US estate-tax exposure on US-listed ETFs (situs rules and treaties vary by country). All funds are accumulating (no dividends to reinvest, no withholding tax friction). RWX→IPRE and IGOV→IGLA are deliberate proxies rather than like-for-like equivalents: IPRE tracks European property (RWX is global ex-US) and IGLA tracks global government bonds including US Treasuries (IGOV excludes the US).

How these numbers are computed

The backtest ranks and measures on the US tickers above, which have the longest clean history; live execution buys their UCITS equivalents.

Curves and metrics are gross: no spread, commission, slippage or tax is deducted. All-in trading costs run roughly 5 to 25 bps per unit of turnover, a drag of about 0.2 to 1.0 CAGR point per year at this turnover.

Sharpe here is (CAGR − 4%) / volatility: a compounded return, minus a fixed 4% risk-free rate, per unit of volatility. Published Sharpe figures usually take the arithmetic average return and the risk-free rate of the day, so the two are not directly comparable.

Win Rate is the share of positive months, not the share of winning trades. Momentum rotation wins less often than that, and makes it up on the size of the winners.