Three independent decision layers, one unified portfolio
Equity Alpha is a benchmark-aware global equity ETF strategy built from three decision layers: ML ETF Selection, Regional Momentum Allocation, and Selective Alpha. Their ETF-level targets are merged into one final portfolio before turnover, transaction costs, and performance are calculated.
ACWI is the primary benchmark. The dashboard reports absolute and benchmark-relative results together, including active return, tracking error, information ratio, relative drawdown, current holdings, and a Preview of the next proposed portfolio.
Different decision paths are combined before implementation
Primary alpha engine
A point-in-time equity ETF universe is ranked using regularized cross-sectional forecasts and Chronos-2. The combined rank signal is translated into holdings through explicit entry and hold rules.
Independent geographic allocation
A separate regional process contributes an independent geographic allocation signal alongside the primary ETF-selection engine.
Targeted Theme / Sector / Style sleeve
A smaller independent path adds focused Theme, Sector, and Style exposures alongside the primary ML engine.
One ETF, one final target weight
Overlapping ETF selections are merged first. Turnover, transaction costs, and realized performance are then calculated from the unified portfolio rather than from three separately traded sleeves.
Completed history, current performance, Official, and Preview are kept distinct
Completed performance uses closed daily history. Current MTD/YTD uses the latest available market closes. Official is the portfolio for the current holding month. Preview is the synchronized next-period candidate portfolio and is never presented as executed.
ACWI anchors the active-risk view
The strategy is evaluated against ACWI using active return, tracking error, information ratio, relative wealth, and relative drawdown alongside absolute return and drawdown. The dashboard can also combine Global Alpha with an ACWI core; that control changes the investor-level risk profile and investable weights, not the underlying model selections.
Live Completed Performance
Updated with each canonical Equity Alpha release. The figures below show completed daily performance for the 100% Global Equity Alpha portfolio against ACWI, using the same completed-history authority as the public dashboard. Current MTD and Preview remain separate operating states and are excluded here.
Cumulative Performance
2019-01-31 → 2026-09-30 · completed history only
Recent Completed Monthly Returns
Latest 12 completed months · 2025-10 → 2026-09 · current MTD excluded
2019-01-31 → 2026-09-30 · 1,926 completed daily returns · daily-first accounting
| Series | Cumulative | CAGR | Vol | Sharpe | MDD |
|---|---|---|---|---|---|
| Global Equity Alpha | 273.75% | 18.83% | 21.45% | 0.91 | -33.10% |
| ACWI benchmark | 162.49% | 13.46% | 18.35% | 0.78 | -33.52% |
A public operating view of the unified portfolio
The live dashboard shows Official and Preview portfolios, ETF target weights, investment-exposure groupings, current MTD/YTD, completed daily performance, ACWI-relative risk, portfolio history, and ML-layer diagnostics. A full-history workbook provides the completed daily and monthly record together with the unified target-weight history.
The public view is designed to explain portfolio construction, operating state, and portfolio-level evidence; proprietary implementation details are not published.
Methods used in this system
These companion articles explain methods used directly in the system or needed to interpret its validation and outputs.
Look-Ahead Bias and Data Leakage
Explains the look-ahead, survivorship, and data-leakage controls required when Equity Alpha reconstructs point-in-time ETF opportunity sets and model inputs.
Portfolio Backtesting and Rebalancing
Defines the monthly signal, execution, holding-period, and weight-drift accounting needed to interpret Equity Alpha portfolio paths.
Turnover, Transaction Costs, and Net Performance
Explains turnover and transaction-cost accounting used when ranked selections, entry rules, and hold buffers are translated into net portfolio performance.
Momentum and Trend Signals
Defines momentum and trend concepts used by the independent regional allocation process and momentum-based components of the selective-alpha decision path.
Strategy Similarity and Decision Diversification
Provides the decision-diversification framework used to combine ML selection, regional allocation, and selective alpha without treating one historical optimum as the only valid decision path.
Benchmark-Relative Portfolio Evaluation
Defines active return, tracking error, information ratio, relative wealth, and relative drawdown used to evaluate Equity Alpha against ACWI.
Regularized Linear Models: Ridge, Lasso, and Elastic Net
Defines the Ridge and Elastic Net model family used inside the REX2 regularized cross-sectional forecasting lane.
Pretrained Time-Series Foundation Models
Defines the evaluation boundary for Chronos-2 and other pretrained time-series foundation models, separating inference-time information from pretrained knowledge and historical-evaluation claims.
Point-in-Time Dynamic Universe Construction
Defines point-in-time dynamic-universe construction, including listing history, liquidity/history gates, survivorship control, exposure deduplication, and controlled expansion of the model-ready candidate pool.
Cross-Sectional Ranking, Top-K Selection, and Hold Buffers
Explains how cross-sectional forecasts become ranks, how Top-K entry and hold buffers create path-dependent selections, and why rank diagnostics are not expected returns or probabilities.