Vigilant Asset Allocation (VAA-G4)
Developed by Keller & Keuning · Breadth Momentum · High Risk
Vigilant Asset Allocation was published by Wouter Keller and Jan Willem Keuning in their 2017 SSRN paper (#3002624), introducing a breadth-based crash protection mechanism that represented a significant departure from the trend-filtering approaches that had dominated tactical allocation research. Rather than evaluating whether individual assets were above or below their moving averages, VAA monitors the collective momentum health of a small asset universe and triggers a full portfolio rotation when any single member weakens — an aggressive sentinel approach that prioritizes capital preservation above all else.
Keller, a former professor at VU University Amsterdam with extensive publications in quantitative finance, designed VAA's weighted momentum composite to capture trend information across multiple timeframes simultaneously while giving disproportionate weight to recent price action. The formula — twelve times the one-month return, plus four times the three-month return, plus two times the six-month return, plus one times the twelve-month return — makes the signal highly responsive to current conditions while still incorporating longer-term trend context. This emphasis on recency allows the strategy to detect momentum deterioration earlier than single-period lookbacks, at the cost of increased sensitivity to short-term noise.
The G4 variant applies the most aggressive crash protection in the Keller strategy family. It monitors just four core assets — US stocks, international developed stocks, emerging markets, and aggregate bonds — and if any single one shows negative weighted momentum, the entire portfolio switches to the best-performing defensive asset. This any-one-negative rule means the strategy treats weakness in any corner of the global investment landscape as sufficient reason to exit risk entirely. The rationale is grounded in the empirical observation that market crises rarely arrive without advance warning in at least one asset class.
When all four assets show positive momentum, the portfolio goes all-in on the single highest-scoring asset, creating a high-conviction concentrated position. This binary approach — maximum offense or maximum defense — produces a strategy that spends roughly half its time in defensive positioning due to the hair-trigger sensitivity of the breadth rule. Despite the frequent switching, the effectiveness of the breadth-based crash detection signal has historically delivered strong risk-adjusted returns.
How It Works
The Weighted Momentum Composite (13612W)
Each month, four assets — US stocks (SPY), international developed stocks (EFA), emerging markets (EEM), and US aggregate bonds (AGG) — are scored using the weighted momentum formula: 12 times the one-month return, plus 4 times the three-month return, plus 2 times the six-month return, plus 1 times the twelve-month return. This weighting structure multiplies the most recent monthly return by twelve, making the signal highly responsive to current market conditions. The declining weights on longer-term returns provide trend context without allowing stale information to dominate.
The result is a composite that reacts faster than a standard twelve-month momentum calculation but retains awareness of the broader trend trajectory. A sharp one-month decline carries substantial weight in the score, potentially turning it negative even if longer-term returns remain positive. This design reflects Keller's view that the earliest signs of momentum deterioration are the most actionable — waiting for confirmation across longer timeframes reduces the speed advantage that makes the protection mechanism effective.
The Any-One-Negative Breadth Trigger
The protection mechanism is deliberately aggressive: if any single asset in the four-member universe produces a negative 13612W score, the entire portfolio switches to defensive mode. Weakness in emerging markets alone, or in bonds alone, is sufficient to trigger a full exit from risk assets — even if the other three members show strong positive momentum.
This design reflects the empirical observation that market crises rarely arrive without advance warning in at least one asset class. Emerging markets often weaken first during global risk-off episodes due to capital flight and currency pressure. Aggregate bonds can deteriorate ahead of equity weakness when rising rates signal tightening financial conditions. By responding to the earliest available warning signal rather than waiting for confirmation across multiple assets, the strategy accepts a higher rate of false alarms in exchange for earlier exits when genuine danger materializes.
Asset Selection: Offense and Defense
During risk-on periods — when all four assets show positive momentum — the portfolio invests 100% in the single highest-scoring offensive asset. This concentrated positioning maximizes exposure to the strongest current trend, producing outsized returns when the ranking is correct but amplifying the impact of any ranking error.
During defensive periods, the portfolio rotates into the best-performing member of a three-asset safe-haven universe: short-term Treasuries (SHY), intermediate-term Treasuries (IEF), or investment-grade corporate bonds (LQD). The defensive selection also uses 13612W scoring, directing capital to whichever safe-haven asset is trending most favorably. This adaptive defensive positioning means the strategy can favor duration when rates are falling or credit when spreads are compressing, rather than defaulting to a single static defensive asset.
The combination of concentrated offense, aggressive protection triggers, and adaptive defense creates a strategy with very high turnover — frequent transitions between offensive and defensive modes, and rotation within each mode as rankings change. This turnover is the cost of the strategy's responsiveness, and must be weighed against the protection benefits when evaluating suitability for different account types.
Explore Vigilant Asset Allocation (VAA-G4)
See the full backtest across 18 years of market data, or run your own what-if scenarios by adjusting all research parameters.