Paper: arXiv 2406.01199

Authors: Alexandre V. Antonov, Koushik Balasubramanian, Alexander Lipton, Marcos Lopez de Prado

Abstract

In this article, a geometric approach to incorporating investor views in portfolio construction is presented. In particular, the proposed approach utilizes the notion of generalized Wasserstein barycenter (GWB) to combine the statistical information about asset returns with investor views to obtain an updated estimate of the asset drifts and covariance, which are then fed into a mean-variance optimizer as inputs. Quantitative comparisons of the proposed geometric approach with the conventional Black-Litterman model (and a closely related variant) are presented. The proposed geometric approach provides investors with more flexibility in specifying their confidence in their views than conventional Black-Litterman model-based approaches. The geometric approach also rewards the investors more for making correct decisions than conventional BL based approaches. We provide empirical and theoretical justifications for our claim.

Complexity vs Empirical Score

  • Math Complexity: 9.0/10
  • Empirical Rigor: 3.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper introduces complex optimal transport theory (Wasserstein barycenters, McCann interpolants) with significant mathematical derivation, while the empirical section describes a backtesting methodology without presenting code or heavy implementation details.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Develop a geometric approach<br>to incorporate investor views<br>with more flexibility than BL model"] --> B["Data & Inputs"]
  
  subgraph B ["Data & Inputs"]
      B1["Asset Return Data"]
      B2["Statistical Estimates<br>Drifts & Covariance"]
      B3["Investor Views<br>Return Expectations"]
      B4["Confidence Level<br>Flexible/Parameterized"]
  end
  
  B --> C["Core Methodology<br>Generalized Wasserstein Barycenter"]
  
  C --> D["Computational Process<br>Combine Statistical Info + Views"]
  D --> E["Output: Updated<br>Drift & Covariance"]
  E --> F["Mean-Variance<br>Optimizer"]
  
  F --> G["Key Findings/Outcomes"]
  
  subgraph G ["Key Findings/Outcomes"]
      G1["Geometric Approach<br>provides more flexibility<br>in specifying confidence"]
      G2["Rewards investors more<br>for correct decisions<br>vs. BL model"]
      G3["Validated by<br>Empirical & Theoretical<br>justifications"]
  end

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 geometric approach that uses the generalized Wasserstein barycenter to incorporate investor views into portfolio construction.

Source sentence
In this article, a geometric approach to incorporating investor views in portfolio construction is presented.

Assumes 5

  • Observable asset returns are assumed to follow a Gaussian distribution with mean µR and covariance CR. statistical
    Source sentence
    The observable asset returns (⃗R) are assumed to follow a Gaussian distribution centered around a mean (⃗µR) and the covariance of the returns is denoted by CR.
  • The estimate of the drift bµR is assumed to be normally distributed with covariance Cd. statistical
    Source sentence
    The estimate of the drift b⃗µR is assumed to be normally distributed with covariance (Cd): b⃗µR =⃗µd +⃗ϵd, where⃗ϵd ∼N(⃗0Na, Cd),
  • Risk‑free rate is set to zero in the mean‑variance optimization. model
    Source sentence
    In the rest of this article, we will assume rf = 0.
  • Quarterly rebalancing schedule is used for all allocation methodologies. execution
    Source sentence
    We use a quarterly rebalancing schedule for all the four allocation methodologies.
Show 1 more
  • Investor confidence parameter t = λ/(1+λ) can be tuned dynamically or set as a hyper‑parameter. other
    Source sentence
    In principle, t can be tuned dynamically or determined through a hyperparameter tuning methodology.

Data

Backtest sample: S&P 500 stocks · equities · daily

  • S&P 500 constituents market-prices unknown
    around fifteen years of data · Na = 50 stocks at random from >350 S&P 500 constituents
    Source sentence
    For the purpose of this study we choose the stocks which are the current constituents of S & P 500 having around fifteen years of data as the universe U. This has over 350 stocks out of which we choose Na = 50 stocks at random.

Recognised sources: Yahoo Finance

Costs

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

Source sentence
We have not included the impacts of transaction costs, holding or borrowing costs, slippage, etc., in our present analysis as the primary goal of this study is to compare the efficacy of the drift and covariance corrections.

Finds 2

Reported performance (GWB-II (High confidence)): Sharpe 1.52 (benchmark 0.54)

  • The geometric approaches (GWBI and GWBII) achieve higher Sharpe ratios than the conventional Black‑Litterman models when investor views are correct. performance
    Source sentence
    It can be inferred from the location of the peaks of the histograms that the geometric approaches outperform the BL models.
  • The geometric approach provides investors with more flexibility to specify confidence in their views compared to conventional BL models. theoretical
    Source sentence
    The approach presented in this paper provides an investor the flexibility to specify the confidence in the form of a parameter that does not exist in the conventional BL models.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 13 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 by independent authors)

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 (22)

  1. Mitochondrial DNA and genetic disease. (1992) — link
  2. Convex optimization theory (2010) — link
  3. A class of Wasserstein metrics for probability distributions. (1984) — link
  4. The distance between two random vectors with given dispersion matrices (1982) — link
  5. The Fréchet distance between multivariate normal distributions (1982) — link
  6. Wasserstein geometry of Gaussian measures (2011) — link
  7. A Convexity Principle for Interacting Gases (1997) — link
  8. CVXPY: A Python-Embedded Modeling Language for Convex Optimization. (2016)
and 14 more
  1. Convex Optimization (2004) — link
  2. On the Bayesian interpretation of Black–Litterman (2016) — link
  3. Incorporating Trading Strategies in the Black-Litterman Framework (2006) — link
  4. The Black-Litterman Approach: Original Model and Extensions (2008) — link
  5. A rewriting system for convex optimization problems (2018) — link
  6. On the Bures–Wasserstein distance between positive definite matrices (2018) — link
  7. CVXPY: A Python-Embedded Modeling Language for Convex Optimization (2016) — link
  8. Bayesian methods in finance (2008)
  9. Advances in Financial Machine Learning (2018)
  10. Incorporating Trading Strategies in the Black-Litterman Framework (2006) — link
  11. Generalized Wasserstein barycenters between probability measures living on different subspaces (2023) — link
  12. Measure Theory (2007) — link
  13. Black-Litterman End-to-End (2023) — link
  14. Underperformance of Concentrated Stock Positions (2023) — link

0 of 22 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.