What Do Additional Models Add to a Global Equity Portfolio?
Overview
More models do not automatically produce a better equity portfolio. This research asks what regional and selective allocations contribute after their positions are combined with a machine-learning ETF selection engine and evaluated against ACWI.
The study examines one investable global equity ETF portfolio, not three separate model accounts. Overlapping ETF targets are consolidated before turnover, transaction costs, and benchmark-relative performance are measured. It tests the value of additional decision layers through their effect on the portfolio an investor would actually hold.
Research Question
Key Findings
Published Evidence
The common daily reconstructed evaluation runs from 31 January 2019 to 31 August 2026. It uses one ACWI benchmark and applies portfolio-level transaction costs after all three components' ETF targets are combined.
Selected published, common-accounting comparisons. The full window ends in August 2026; the later diagnostic covers November 2024 through August 2026.
| Configuration | Window | CAGR | TE | IR | Relative DD |
|---|---|---|---|---|---|
| Primary only | Full | 19.6% | 7.0% | 0.816 | −11.6% |
| + Regional | Full | 18.9% | 6.1% | 0.821 | −9.9% |
| + Selective | Full | 18.9% | 6.2% | 0.819 | −8.1% |
| Primary only | Later | 26.3% | 7.2% | 0.690 | −8.4% |
| + Regional | Later | 25.5% | 6.2% | 0.682 | −7.2% |
| + Selective | Later | 24.7% | 6.2% | 0.582 | −7.1% |
Primary selection alone compounds faster, while Regional allocation reduces tracking error and relative drawdown at the cost of some active return. Adding Selective Alpha further improves full-period relative drawdown, but the later-period information ratio declines. The paper therefore evaluates incremental contribution as a return–risk trade-off, not a presumption that every layer adds alpha.
Portfolio Architecture
ML ETF Selection
The main return-seeking engine ranks historically eligible global equity ETFs, with a separate static-universe anchor.
Regional Allocation
A smaller, benchmark-aware allocation layer adjusts the geographic and U.S.-growth exposure of the unified portfolio.
Selective Alpha
A limited allocation to theme, sector, style, and factor ETFs combines longer-horizon momentum and separate ranking signals.
The published configuration assigns 75% to Primary selection, 20% to Regional allocation, and 5% to Selective Alpha. These are capital allocations to the three components; their internal ranking and portfolio weights are different quantities. ETF positions are merged by ticker before the final portfolio is traded and evaluated.
Research and Operational System
This page describes the published research record. The Equity Alpha System page explains the operating portfolio architecture and presents current completed-history evidence. Its live dashboard separately identifies Official holdings, current open-month performance, and next-period Preview positions.
Methods used in this research
These companion articles explain methods used directly in the study or needed to interpret its design and evidence.
Look-Ahead Bias and Data Leakage
Explains how timing, data construction, and model-development choices can let future information enter a historical forecasting exercise.
Portfolio Backtesting and Rebalancing
Explains the portfolio accounting, weight drift, monthly rebalancing, turnover conventions, and next-period timing needed to interpret the portfolio paths used in this research.
Turnover, Transaction Costs, and Net Performance
Explains drift-aware turnover, transaction-cost conventions, and the mapping from before-cost portfolio performance to performance net of modeled transaction costs.
Momentum and Trend Signals
Explains how momentum and trend signals are defined, timed, and stress-tested across lookback choices before they are translated into portfolio decisions.
Strategy Similarity and Decision Diversification
Explains how return similarity differs from similarity in what strategies hold, when they change, and how strongly they change risk exposure.
Benchmark-Relative Portfolio Evaluation
Explains active return, tracking error, information ratio, relative wealth, and benchmark-relative drawdown as distinct portfolio-evaluation quantities.
Regularized Linear Models: Ridge, Lasso, and Elastic Net
Explains Ridge, Lasso, and Elastic Net with explicit scaling, penalty, correlated-predictor, and leakage-aware tuning conventions for predictive modeling.
Pretrained Time-Series Foundation Models
Explains pretrained time-series forecasting models by separating inference-time data, pretraining knowledge, zero-shot use, covariate timing, and fair historical evaluation claims.
Point-in-Time Dynamic Universe Construction
Explains how an investable universe can evolve through time without survivorship bias, including listing history, eligibility gates, exposure deduplication, representative selection, and controlled candidate-pool expansion.
Cross-Sectional Ranking, Top-K Selection, and Hold Buffers
Explains the translation from cross-sectional scores to ranks and portfolio selections, including normalized rank strength, Top-K entry, incumbent hold buffers, turnover trade-offs, and path dependence.