---
title: Synthetic data for portfolios: a throw of the dice will never abolish chance
url: https://www.ml-quant.com/papers/arxiv/2501.03993/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-09-26
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: arXiv:2501.03993
source_url: http://arxiv.org/abs/2501.03993v1
featured: 2025-01-08
citations: 10
topic: Portfolio & Allocation
---


# Synthetic data for portfolios: a throw of the dice will never abolish chance

The paper highlights the shortcomings of generative models in finance, suggesting a method for generating multivariate returns that meets standard evaluations and aligns with observed asset returns.

- Source: http://arxiv.org/abs/2501.03993v1
- Identifier: arXiv:2501.03993
- Released: 2025-01-07
- First featured: Quant Letter No. 81 (2025-01-08): https://www.ml-quant.com/issues/2025-01-08/
- Citations (Semantic Scholar): 10
- Published in: Quantitative Finance
- Topic: Portfolio & Allocation

## Related

- [Improving Portfolio Construction with Synthetic Time Series Data](https://www.ml-quant.com/papers/ssrn/4809931/): A deep generative model can create high-quality, correlation-aware synthetic financial time series, useful in a GAN-assisted risk-based portfolio construction method.
- [Dynamic data generation and dynamic portfolio selection: an application of a score-based diffusion model](https://www.ml-quant.com/papers/arxiv/2507.09916/): A model-free approach using generative diffusion models and a policy gradient algorithm has been suggested for dynamic mean-variance portfolio selection, outperforming the Markowitz portfolio and S&P 500.
- [Improved Data Generation for Enhanced Asset Allocation: A Synthetic Dataset Approach for the Fixed Income Universe](https://www.ml-quant.com/papers/arxiv/2311.16004/): The study presents a method for creating synthetic datasets to evaluate asset allocation methods and build portfolios within the fixed income universe.
- [Quantifying Credit Portfolio sensitivity to asset correlations with interpretable generative neural networks](https://www.ml-quant.com/papers/arxiv/2309.08652/): A unique method using Variational Autoencoders has been suggested to measure credit portfolio Value-at-Risk sensitivity to asset correlations, providing a clearer latent space representation.
- [A monotone numerical integration method for mean-variance portfolio optimization under jump-diffusion models](https://www.ml-quant.com/papers/arxiv/2309.05977/): The research introduces an efficient numerical integration method for portfolio optimization, demonstrating its computational efficiency and accuracy, and its convergence to the unique solution of the optimization problem.
- [Generative Meta-Learning Robust Quality-Diversity Portfolio](https://www.ml-quant.com/papers/arxiv/2307.07811/): The paper suggests a meta-learning method for creating a robust portfolio ensemble using a deep generative model, which balances sub-portfolio performance and correlation minimization, making it resilient to systematic shocks.