Investment Research Series

Diversify the Decisions, Not Just the Assets

A Practical Architecture for Dynamic Asset Allocation When Strategy Choice Is Uncertain

Sungkyu LeeGraduate School of Computing, Yonsei UniversityPublic Working PaperSSRN 7251518August 2026

Open SSRN Record Replication GitHub Research Dashboard

Overview

ADAA treats strategy selection as a source of uncertainty rather than assuming that one allocation rule can be identified as best in advance. It asks whether diversification should extend beyond assets to the decision rules that determine portfolio exposures.

The framework compares several dynamic allocation sleeves by what they choose, when they change, and how strongly they move toward defensive exposure. The goal is to identify genuine differences in decision behavior and to evaluate whether combining those differences can reduce dependence on a single rule.

Research Question

Can a dynamic asset-allocation portfolio reduce dependence on choosing the single best rule in advance by diversifying across genuinely different decision processes?

Key Findings

Decision diversity is not guaranteed by strategy count. Several strategies can share the same underlying timing or momentum logic even when their names and holdings differ.

The historical ADAA sleeves disagree in economically meaningful ways, but unevenly. The analysis makes redundancy visible instead of assuming diversification from labels.

The practitioner weights sit inside a broad near-optimal region rather than at a unique optimum. The evidence favors robustness to weight uncertainty over precise weight estimates that are not supported by the sample.

ADAA is not a return-maximizing claim. The case for ADAA rests on decision diversification, implementation, and risk-return trade-offs.

Selected Evidence

Selected summary statistics from the historical simulation reported in Public Working Paper v1.25. SSRN 7251518.

10.80%

ADAA practitioner gross CAGR

8.97%

Annualized volatility

1.05

BIL-excess Sharpe

−10.34%

Maximum drawdown

These statistics and figures correspond to the cited public research version. Live implementation views are separate and may change over time.

Selected Exhibits

Three paper-aligned figures summarize the central evidence. Click any figure to inspect the full-resolution public research image.

ADAA figure comparing return correlation with disagreement in decision timing across the current strategy sleeves. — Public ADAA research · Figure 1
Return Correlation vs. Decision Timing
Return correlation does not fully describe decision diversification. The current sleeves can produce correlated return paths while still changing exposures at materially different times.
Open full-size figure ↗

ADAA decision-fingerprint figure comparing what the sleeves hold, when they change, and how much risk they take. — Public ADAA research · Figure 2
Decision Fingerprints: What, When, and How Much
The five sleeves are compared across three distinct dimensions: what they hold, when they change, and how strongly they alter risk exposure.
Open full-size figure ↗

ADAA robustness figure showing a broad high-performing weight region and instability of the exact ex-post optimum. — Public ADAA research · Figure 5
Broad Plateau, Unstable Optimum
The robustness exercise emphasizes a broad high-performing region rather than a single ex-post peak, supporting moderate practitioner weights instead of precision optimization.
Open full-size figure ↗

Explore the full ADAA Research Dashboard ↗

Methodology

01 · WHAT

Portfolio decisions

Compare which exposures the rules choose and how far their target weights differ.

02 · WHEN

Decision clocks

Measure whether different rules actually change their targets at the same times.

03 · HOW MUCH

Risk response

Track the amount of defensive exposure and the scale of allocation changes.

Reproducibility

Public code and reproducibility materials are maintained in the project repository. Additional replication or archival links appear here when they are part of the public research release.

Public code

GitHub Repository

Research code and public reproducibility material.

Open GitHub ↗

Replication

Replication Package

Current public replication release: v1.1.2.

Open release ↗

Archival record

Release / DOI

Archived public replication release with persistent DOI.

Open DOI ↗

Research and Implementation

ADAA can be viewed through two interfaces with distinct purposes: one documents versioned research evidence, while the other supports live implementation and monitoring.

Public research evidence

ADAA Research Dashboard

Paper-aligned, versioned research evidence, selected figures, and links to the corresponding public research materials.

Open Research Dashboard

Implementation

ADAA Live / Operational Dashboard

Current implementation and monitoring context. This interface can change with live conditions and is not the versioned research record.

Open Live / Operational Dashboard

Version boundary: research results on this page correspond to the cited public paper version; live operational views may change independently.

Citation

Lee, S. (2026). Diversify the Decisions, Not Just the Assets: A Practical Architecture for Dynamic Asset Allocation When Strategy Choice Is Uncertain. Public Working Paper v1.25. SSRN 7251518.