Adaptive Asset Allocation (AAA)
Developed by Adam Butler · Risk-Based Momentum · Med Risk
Adaptive Asset Allocation was developed by Adam Butler, Michael Philbrick, and Rodrigo Gordillo at ReSolve Asset Management, published in their 2012 SSRN paper (#2328254) and later expanded in their book Adaptive Asset Allocation: Dynamic Global Portfolios to Profit in Good Times and Bad. The strategy represents a fundamentally different philosophy from the trend-following and momentum-switching approaches that dominate the tactical allocation landscape: rather than alternating between risk-on and risk-off modes, AAA embeds risk management directly into the portfolio construction process itself.
Butler's research background in risk parity and portfolio optimization led him to recognize that the conventional momentum approach — select the strongest assets, then equal-weight them — ignores a critical dimension of portfolio risk. Two assets with identical momentum but different volatilities contribute very different amounts of risk to the overall portfolio. Two assets with identical momentum but high mutual correlation provide far less diversification benefit than two with low correlation. Standard momentum strategies make no distinction between these scenarios, treating all selected assets as interchangeable building blocks.
AAA addresses this gap by adding a minimum variance optimization step after momentum selection. The strategy selects the top five assets from a globally diversified ten-asset universe based on six-month momentum, then uses a covariance matrix estimated from recent daily returns to calculate the combination of weights that minimizes expected portfolio volatility. This dual filter — momentum for selection, optimization for weighting — creates portfolios that capture trending assets while actively managing how much risk each position contributes to the total.
The ten-asset universe spans US, European, Japanese, and emerging market equities alongside real estate, commodities, gold, and bonds. This breadth ensures that the optimization engine has access to genuinely diverse building blocks across different asset classes, geographies, and risk characteristics. The optimizer's effectiveness is directly proportional to the diversity of its inputs — a broader, less correlated universe produces portfolios with lower realized volatility than a narrow one.
How It Works
Momentum Screening
Each month, all ten assets — US equities (SPY), European equities (VGK), Japanese equities (EWJ), emerging markets (EEM), US real estate (VNQ), international real estate (RWX), intermediate Treasuries (IEF), long-term Treasuries (TLT), commodities (DBC), and gold (GLD) — are ranked by their trailing six-month total return. The top five by momentum score advance to the optimization stage.
The six-month lookback represents a middle ground between faster signals like ADM's blended one-three-six composite and slower approaches like GEM's twelve-month window. Academic research on the momentum effect shows that it peaks at horizons of three to twelve months across asset classes, with six months capturing the core of the phenomenon while providing enough responsiveness to avoid holding assets deep into established downtrends.
Minimum Variance Optimization
The five selected assets are passed into a long-only minimum variance optimizer. Rather than giving each winner an equal twenty percent weight, the optimizer calculates the combination of weights that produces the lowest expected portfolio volatility based on a covariance matrix estimated from the trailing 126 trading days (approximately six months) of daily returns.
The optimizer naturally produces several desirable properties without requiring any explicit rules. Assets with high individual volatility receive lower weights because they contribute more variance per unit of allocation. Assets highly correlated with other holdings receive lower combined weight because their diversification benefit is limited. Assets with low or negative correlation to other holdings receive higher weight because they reduce portfolio-level risk. The result is a portfolio inherently tilted toward more stable, diversifying positions — a risk-managed allocation that emerges from the optimization math rather than from subjective judgment about which assets deserve more emphasis.
Continuous Adaptation Without Defensive Switching
AAA has no explicit defensive mode, no cash switching mechanism, and no binary trigger. The portfolio is always fully invested in its top five momentum-ranked and variance-optimized positions. Protection emerges organically from two complementary sources working simultaneously.
First, the breadth of the universe — which includes bonds, gold, and commodities alongside equities — means that during equity bear markets, non-equity assets naturally rise in the momentum rankings and begin displacing equity holdings. Second, the variance-minimizing allocation simultaneously increases the weight of lower-volatility, lower-correlation assets during turbulent markets when equity correlations spike and volatilities expand. This dual effect — momentum selection rotating toward safe havens and optimization amplifying their weight — creates a natural defensive rotation that operates without any explicit threshold or signal.
During the 2008 financial crisis, this mechanism shifted the portfolio progressively toward bonds and gold as equity momentum deteriorated and their volatilities increased relative to non-equity alternatives. The transition was gradual rather than binary — equities lost momentum ranking over several months while the optimizer simultaneously reduced their remaining weights as cross-asset correlations spiked. This smooth adaptation contrasts with the sharp, decisive transitions of binary momentum strategies, producing a different risk-return profile: fewer large drawdowns but slower response to rapid market changes.
Explore Adaptive Asset Allocation (AAA)
See the full backtest across 18 years of market data, or run your own what-if scenarios by adjusting all research parameters.