Paper: arXiv 2311.04475

Authors: Fanyu Zhao

Abstract

This paper presents a portfolio construction process, including mainly two parts, Factors Selection and Weight Allocations. For the factors selection part, We have chosen 20 factors by considering three aspects, the global market, different assets class, and stock idiosyncratic characteristics. Each factor is proxied by a corresponding ETF. Then, we would apply several weight allocation methods to those factors, including two fixed weight allocation methods, three optimisation methods, and a Black-Litterman model. In addition, we would also fit a Deep Learning model for generating views periodically and incorporating views with the prior to achieve dynamically updated weights by using the Black-Litterman model. In the end, the robustness checking shows how weights change with respect to time evolving and variance increasing. Results using shrinkage variance are provided to alleviate the impacts of representativeness of historical data, but there sadly has little impact. Overall, the model by using the Deep Learning plus Black-Litterman model results outperform the portfolio by other weight allocation schemes, even though further improvement and robustness checking should be performed.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper employs advanced mathematical frameworks including the Black-Litterman model, mean-variance optimization, and deep learning for view generation, with substantial notation and matrix algebra. However, it relies heavily on theoretical models and backtesting with limited discussion of transaction costs, data cleaning, or robust out-of-sample validation, focusing more on methodology than implementation details.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Portfolio Construction via<br>Black-Litterman & Factors"] --> B["Factor Selection<br>20 Factors via 3 Aspects<br>(Market, Asset Class, Idio.)"]
  B --> C["Data Inputs:<br>ETF Proxies &<br>Historical Market Data"]
  C --> D{"Weight Allocation<br>Methods"}
  D --> E["Fixed Allocation<br>(Equal/Risk Parity)"]
  D --> F["Optimization Methods<br>(Mean-Variance, etc.)"]
  D --> G["Black-Litterman Model<br>Prior + Views"]
  G --> H["Deep Learning Model<br>Generates Dynamic Views"]
  H --> G
  E & F & G --> I["Robustness Checks<br>Time Evolution &<br>Variance Sensitivity"]
  I --> J["Key Findings:<br>DL + BL Model Outperforms<br>Others; Shrinkage<br>Variance Minimal Impact"]

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 Deep Learning model to generate periodic views and integrates them with a Black‑Litterman framework to produce dynamically updated portfolio weights.

Source sentence
In addition, we would also fit a Deep Learning model for generating views periodically and incorporating views with the prior to achieve dynamically updated weights by using the Black-Litterman model.

Assumes 3

  • Weighting factors by market capitalisation assumes the market is already in equilibrium. market
    Source sentence
    using the market cap to weight factors is based on the assumption that the market is already in equilibrium.
  • Views are independent of each other, making the view‑uncertainty matrix diagonal. model
    Source sentence
    By assuming views are independent to each other, the variance‑covariance and correlation between views are all zero, so that Ω is a diagonal matrix.
  • All time‑series data for factors are retrieved from Yahoo!Finance. data
    Source sentence
    All the time series data are retrieved from Yahoo!Finance.

Data

  • ETF price and market‑cap data for 20 factors · Yahoo!Finance market-prices public
    20 factor ETFs (US and China equities, bonds, commodities)
    Source sentence
    All the time series data are retrieved from Yahoo!Finance.
  • Market Cap Weights Table · Yahoo!Finance (manual collection) fundamentals public
    20 factor ETFs
    Source sentence
    Table 2: Market Cap Weights ... Asset Class Market Cap Weight (Billion US Dollars) us_growth 0.279375 ...

Recognised sources: Yahoo Finance

Costs

Not modeled A trading or execution result is reported without costs.

Finds 3

  • The Deep Learning plus Black‑Litterman model outperforms portfolios built with other weight‑allocation schemes. performance
    Source sentence
    Overall, the model by using the Deep Learning plus Black-Litterman model results outperform the portfolio by other weight allocation schemes, even though further improvement and robustness checking should be performed.
  • Applying shrinkage variance has little impact on alleviating the representativeness issue of historical data. negative
    Source sentence
    Results using shrinkage variance are provided to alleviate the impacts of representativeness of historical data, but there sadly has little impact.
  • Robustness checks show that factor weights change over time and with increasing variance. statistical
    Source sentence
    the robustness checking shows how weights change with respect to time evolving and variance increasing.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 10 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 (11)

  1. The Black–Litterman model for active portfolio management (2009)
  2. The Black-Litterman model in detail (2014)
  3. A step-by-step guide to the Black-Litterman model: Incorporating user-specified confidence levels (2007)
  4. Active portfolio management (1999)
  5. Active fixed-income portfolio management using the Black-Litterman model (2009)
  6. A demystification of the Black–Litterman model: Managing quantitative and traditional portfolio construction (2000) — link
  7. Size, value, and momentum in international stock returns (2012)
  8. Common Risk Factors in the Returns on Stocks and Bonds (1993)
and 3 more
  1. The intuition behind Black-Litterman model portfolios (2002)
  2. Improved estimation of the covariance matrix of stock returns with an application to portfolio selection (2003)
  3. A Practitioner’s Guide to Multi-Factor Portfolio Construction (2020)

0 of 11 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.