Paper: arXiv 2504.02841
Authors: Jinhui Li, Wenjia Xie, Luis Seco
Abstract
This study introduces a dynamic investment framework to enhance portfolio management in volatile markets, offering clear advantages over traditional static strategies. Evaluates four conventional approaches : equal weighted, minimum variance, maximum diversification, and equal risk contribution under dynamic conditions. Using K means clustering, the market is segmented into ten volatility-based states, with transitions forecasted by a Bayesian Markov switching model employing Dirichlet priors and Gibbs sampling. This enables real-time asset allocation adjustments. Tested across two asset sets, the dynamic portfolio consistently achieves significantly higher risk-adjusted returns and substantially higher total returns, outperforming most static methods. By integrating classical optimization with machine learning and Bayesian techniques, this research provides a robust strategy for optimizing investment outcomes in unpredictable market environments.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs not modelled; no out-of-sample test; see penalty flags)
- Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.0)
Why this score: The paper employs advanced mathematics including K-means clustering, Bayesian Markov switching models with Dirichlet priors and Gibbs sampling, and dynamic optimization. It demonstrates high empirical rigor through rigorous backtesting across two asset sets, validation of clustering, and clear performance metrics (Sharpe ratio, volatility, total returns) comparing dynamic vs. static strategies.
Research Flowchart
flowchart TD A["Research Goal<br>Dynamic Investment Strategy Optimization"] --> B["Data<br>Multi-Asset Portfolio Returns"] B --> C["K-Means Clustering<br>Market Volatility States (10)"] C --> D["Bayesian Markov Switching<br>Forecast Market Transitions"] D --> E["Optimization Engine<br>Dynamic Portfolio Allocation"] E --> F["Evaluation<br>Risk-Adjusted & Total Returns"] F --> G["Outcome<br>Outperforms Static Strategies"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a dynamic investment framework that uses K‑means market state classification and a Bayesian Markov transition matrix with Dirichlet priors and Gibbs sampling to adjust portfolio weights in real time.
Source sentence
This study introduces a dynamic investment framework to enhance portfolio management in volatile markets, offering clear advantages over traditional static strategies.
Assumes 6
- Transition probabilities follow a Dirichlet prior distribution. model
Source sentence
When updating the posterior row-by-row with the Dirichlet distribution Dirichlet(αi+ Pk j=1 N j i), where αi is the prior vector for state i, ...
- Gibbs sampling is used for posterior updating instead of a fixed Dirichlet matrix. model
Source sentence
This sequential updating process is the primary reason we apply Gibbs sampling rather than assuming the transition matrix follows a fixed Dirichlet distribution.
- The transition matrix is updated once per period (e.g., every three or six months). execution
Source sentence
In practice, updates to the transition matrix N(k) ij will occur once per period, such as every three or six months.
- Historical market behavior will repeat in the future. data
Source sentence
The study’s reliance on historical data assumes that past market behavior will repeat, which may not always hold true.
Show 2 more
- Transaction costs and other practical constraints are excluded from the analysis. execution
Source sentence
the exclusion of transaction costs and other practical constraints may impact real‑world performance.
- Market segmentation is based on 22‑day volatility data. data
Source sentence
We apply it to 22‑day volatility data, enabling a granular analysis of how portfolio strategies perform across different market states.
Data
2005 – 2024 · S&P 500 stocks · equities · daily
- First asset set: SPY Top 11 Portfolio market-prices unknown
Source sentence
The cumulative return was derived using the compounded return method over the entire period: 2005 to 2024 for the first asset set
- Second asset set market-prices unknown
Source sentence
The cumulative return was derived using the compounded return method over the entire period:
Costs
Not modeled
Source sentence
the exclusion of transaction costs and other practical constraints may impact real‑world performance.
Finds 5
total return 4910.0%
- For the first set of assets, the dynamic strategy achieved a total return of 4910%, significantly outperforming several static methods. performance
Source sentence
For the first set of assets, it achieved a total return of 4910%, significantly outperforming several static methods.
- For the second set of assets, the dynamic strategy achieved a total return of 5993%, showing superior performance compared to the maximum diversification method, though slightly below equal investment, ERC and Min Var. performance
Source sentence
dynamic strategy achieved a total return of 5993%, showing superior performance compared to the maximum diversification method, although it was slightly less than the equal investment, equal risk contribution and minimum variance methods.
- The dynamic strategy consistently delivered the second highest Sharpe ratio among static methods for the first asset set. performance
Source sentence
the dynamic strategy consistently delivered the second highest Sharpe ratio compared to static methods for the first set of assets
- For the second asset set, the dynamic strategy achieved the second highest Sharpe ratio, underscoring its robustness in risk management. performance
Source sentence
For the second set of assets, the dynamic strategy achieved the second highest Sharpe ratio, underscoring its robustness in managing risk and adapting to state changes.
- Mixing time estimates: first asset SLEM 0.9584 → mixing time 109 steps; second asset SLEM 0.9277 → mixing time 62 steps. statistical
Source sentence
first asset, the calculated SLEM was 0.9584, resulting in an estimated mixing time of 109 steps.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 24 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (9)
- Inference from Iterative Simulation Using Multiple Sequences (1992)
- Markov Chains and Mixing Times (2009)
- Market crises and the 1/N asset-allocation strategy (2013) — link
- Portfolio selection (1952)
- The properties of equally weighted risk contribution portfolios (2010)
- Optimal versus naive diversification: How inefficient is the 1/n portfolio strategy? (2009)
- A new approach to the economic analysis of nonstationary time series and the business cycle (1989)
- In defense of optimization: The fallacy of 1/n (2010)
and 1 more
- Forecasting market turbulences using regime-switching models (2012)
0 of 9 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.