Defensive Adaptive Allocation (KDA)
Developed by Ilya Kipnis · Canary + Adaptive · Med Risk
Defensive Adaptive Allocation was developed by Ilya Kipnis, a quantitative researcher and financial engineer, and published on his Quantitative Strategy TradeR blog. KDA merges two powerful concepts from tactical allocation research — Keller's canary universe architecture for crash detection and Butler's minimum variance optimization for portfolio construction — into a single strategy that aims to be both fast at detecting danger and sophisticated in managing portfolio-level risk during favorable conditions.
The canary mechanism borrowed from DAA uses two sentinel assets — emerging markets (VWO) and aggregate bonds (BND) — scored with the 13612W weighted momentum composite to detect deteriorating market conditions before they reach the main investment universe. When either canary shows negative momentum, the portfolio shifts to intermediate bonds. When both canaries are positive, the strategy proceeds to its asset selection and optimization stage.
What distinguishes KDA from other canary-based strategies is the portfolio construction method applied during risk-on periods. Rather than equal-weighting or score-proportional weighting, KDA selects the top five assets from a ten-asset globally diversified universe by six-month momentum and then applies minimum variance optimization to determine their allocation weights. This optimization step — borrowed from Butler's Adaptive Asset Allocation — considers the full covariance structure among the selected assets, tilting the portfolio toward combinations that minimize expected volatility while maintaining exposure to trending assets.
The marriage of fast canary protection with sophisticated portfolio optimization creates a strategy with an unusual risk profile: decisive binary exits during market stress (driven by the canary gate) combined with nuanced, diversification-aware positioning during favorable conditions (driven by the optimizer). This dual character means the strategy can switch abruptly between a carefully optimized multi-asset portfolio and a simple bond allocation, depending on the canary signal — a behavioral contrast that requires investors to accept both the sophistication and the simplicity as integral parts of the system.
How It Works
Canary-Based Crash Detection
Each month, VWO and BND are scored using the 13612W weighted momentum composite — the same signal used in VAA and DAA, emphasizing recent returns through the 12×R1 + 4×R3 + 2×R6 + R12 formula. If either canary shows a negative score, the portfolio shifts entirely to intermediate bonds (IEF). This binary macro gate provides fast, decisive protection: a single negative canary triggers full defensive positioning, accepting false alarms as the cost of early exits during genuine market crises.
The choice of VWO and BND as canaries reflects the same logic as DAA — emerging markets capture global risk appetite deterioration, while aggregate bonds capture rate and credit stress. Together they provide advance warning across the two primary channels through which market stress typically propagates before reaching US large-cap equities.
Momentum Screening and Optimization
When both canaries are positive, ten assets — US equities (SPY), European stocks (VGK), Japanese stocks (EWJ), emerging markets (EEM), US real estate (VNQ), international real estate (RWX), intermediate bonds (IEF), long-term bonds (TLT), commodities (DBC), and gold (GLD) — are ranked by six-month trailing returns. The top five advance to the optimization stage.
The minimum variance optimizer calculates the long-only portfolio weights that minimize expected volatility based on a trailing covariance matrix estimated from recent daily returns. Assets that are volatile receive lower weights. Assets highly correlated with other holdings receive reduced allocation. Assets that diversify the portfolio receive elevated weights. The result is a risk-aware allocation that captures trending assets while actively managing how much each position contributes to total portfolio risk.
The Dual Character
KDA exhibits two distinctly different behavioral modes. During risk-on periods (when both canaries are positive), the portfolio holds a carefully optimized blend of five assets with weights calibrated to minimize volatility — a sophisticated, multi-asset portfolio that adapts its risk profile as correlations and volatilities change. During risk-off periods (when either canary is negative), the portfolio holds a single-asset bond position — the simplest possible defensive allocation.
This contrast creates a strategy whose complexity varies with market conditions. Favorable markets produce a nuanced, diversification-optimized portfolio. Unfavorable markets produce a blunt defensive position. The transition between these two states is binary and immediate — there is no gradual shift from optimized to defensive. This design reflects a philosophy that crash protection requires speed and decisiveness rather than gradual adjustment, while growth capture benefits from sophistication and nuance.
Explore Defensive Adaptive Allocation (KDA)
See the full backtest across 18 years of market data, or run your own what-if scenarios by adjusting all research parameters.