What if picking winning stocks wasn’t about finding isolated outperformers, but about tracing the invisible web of influence that binds the market together? A recent paper proposes exactly that—building portfolios from the market’s structural core, using a dynamic network of directional dependencies extracted from stock returns.
At the heart of the approach lies a clever pipeline that fuses econometrics, network theory, and forecasting:
- Stocks are modeled in pairs using Vector Autoregression (VAR) over rolling 120-day windows.
- Forecast Error Variance Decomposition (FEVD) quantifies how much each stock influences others, generating a directional dependency matrix.
- This matrix is symmetrized and distilled into a Minimum Spanning Tree (MST)—a sparse, cycle-free map of the market’s backbone.
From this tree, the portfolio selects the top-5 most connected stocks (by degree centrality) in each window—stocks that act as systemic hubs. Then, instead of equal weighting, capital is allocated inversely proportional to each stock’s Value at Risk (VaR) or proportionally to its Sharpe ratio. Stocks with lower downside risk or better risk-adjusted returns receive higher weights.
But the innovation doesn’t stop there. The authors introduce a forecast filter: if a stock’s next-day return, predicted via ARIMA or NNAR (Neural Network Autoregression), is negative, it is excluded from the portfolio for that day. The ensemble strategy, dubbed “AllAgree”, requires consensus among predictors to go long.
Results That Turn Heads
| Strategy | Return (%) |
|---|---|
| Buy & Hold (S&P 500) | 18.12 |
| MST + VaR | 37.03 |
| MST + ARIMA + VaR | 40.71 |
| MST + NNAR + VaR | 74.81 |
| MST + AllAgree + VaR | 85.65 |
These results come from a 365-day trading simulation during 2022–2023. Notably, even a fixed MST portfolio, with no dynamic rebalancing, beats the S&P by over 2x.
One surprising insight: MST-selected stocks remain consistently central over time. Despite daily re-estimation, many top-ranked stocks persist—suggesting they occupy robust positions in the market’s structural core. This creates an elegant hybrid: low turnover like index investing, but higher precision and risk control.
Why This Matters
Traditional portfolio strategies rely heavily on correlation matrices and diversification heuristics. But correlations are symmetric and miss causality. Here, the FEVD-driven MST captures directional influence, reflecting how uncertainty flows through the market. It’s a richer lens for systemic importance.
Also, rather than treat forecasts and risk models as separate modules, this strategy integrates them into one coherent pipeline. A stock isn’t selected unless it is:
- Structurally central in the MST,
- Low risk (VaR) or high reward-to-risk (Sharpe), and
- Predicted to rise tomorrow.
This trifecta offers a compelling upgrade to both passive indexing and high-frequency black-box strategies.
Looking Ahead
The framework is modular and extensible:
- Swap VAR/FEVD with graph neural networks or mutual information for nonlinear influence.
- Expand MST to multi-layer networks with macroeconomic, sentiment, or ETF nodes.
- Incorporate cost-aware trading for real-world deployment.
By treating financial markets as evolving information networks, this paper offers a glimpse into a more structurally aware, risk-sensitive, and interpretable future of portfolio design.
Cognaptus: Automate the Present, Incubate the Future.