Strategy 01 · Long-Term Investing

Investing

A systematic, machine learning–driven portfolio that rebalances monthly — selecting the strongest assets from a diversified universe based on objective signals, without emotion and with strict risk controls.

Academic Foundation

Built on decades of
financial research.

The Smart Investor strategy is grounded in well-documented academic phenomena — each principle backed by decades of empirical research across global markets.

01 · Momentum

Cross-Sectional Momentum

Assets that have outperformed their peers over recent months tend to continue doing so — a persistent anomaly documented across global markets for decades. The strategy ranks assets each month and allocates to the strongest.

02 · Factor Investing

Multi-Factor Model

Systematic risk factors — including momentum, trend, and volatility — have been shown to explain persistent return differences across asset classes. A broad set of such factors forms the foundation of the signal generation process.

03 · Risk Parity

Volatility-Adjusted Weighting

Position sizes are adjusted according to each asset's risk contribution, ensuring that no single holding dominates the portfolio's risk profile — regardless of its expected return.

Backtest Performance · 2005 – 2026

21 years. Consistent alpha.

Multi-asset universe · Benchmark: SPY

+13.47%

CAGR

1.08

Sharpe Ratio

−24.09%

Max Drawdown

+14.14%

Alpha vs SPY

0.56

Calmar Ratio

69.9%

Positive Months

Equity Curve

Portfolio vs Benchmark

Strategy +13.47% p.a.

SPY +11.1% p.a.

Period 2005 – 2026

* Actual backtest results. Past performance is not indicative of future results. All figures are gross of taxes and transaction costs.

Process

How the strategy
works.

Data Collection & Feature Engineering

Price and market data is collected monthly for all assets in the universe. A comprehensive set of quantitative features is derived — covering momentum, trend, volatility, volume, and broader market structure — providing a multi-dimensional view of each asset.

Normalisation & Cross-Sectional Ranking

All features are normalised and ranked cross-sectionally within the universe. This ensures comparability across assets and time periods, and prevents any single metric from dominating the signal — without introducing any look-ahead bias.

Feature Selection

A systematic selection process identifies the most predictive and non-redundant features for the current market environment. Only robust, uncorrelated signals are retained — keeping the model lean and resistant to overfitting.

Model Training & Signal Generation

A machine learning classifier is trained strictly on historical data preceding each signal date. It produces a ranking of all assets — from which the top performers are selected for the coming month's portfolio.

Position Sizing & Rebalancing

Each selected asset receives a position sized according to its risk contribution to the overall portfolio. Positions are rebalanced systematically at month-end, with orders executed automatically at closing prices.