Does Reinforcement Learning Offer a Higher Return Than StaticOptimization? An Analysis for Cryptocurrencies in the Period 2021-2024

Authors

  • Helberte João França Almeida Professor, Dr. Universidade Federal de Santa Catarina
  • Adilson Giovanini Santa Catarina State University image/svg+xml
  • vitor Augusto Ferreira Universidade Federal de Santa Catarina image/svg+xml

DOI:

https://doi.org/10.31501/1dzygs20

Abstract

This work seeks to use Machine Learning techniques, specifically Reinforcement Learning, to optimize resource allocation in a portfolio composed of five large-cap cryptocurrencies: Bitcoin (BTC), Ethereum (ETH), Binance Coin (BNB), Ripple (XRP), and Solana (SOL) in the period of 2021 the 2024. The results indicate that the Reinforcement Learning model outperformed traditional strategies in the bull market, with a cumulative return of 73.26%, however, struggled in the bear market and sideways market, failing to outperform static allocation. Therefore, conclusion is that the model is effective in bull markets but requires adjustments to improve its performance in adverse scenarios.

 

Author Biographies

  • Adilson Giovanini, Santa Catarina State University

    Prof. Dr. Universidade do Estado de Santa Catarina

  • vitor Augusto Ferreira, Universidade Federal de Santa Catarina

    Graduado, Universidade Federal de Santa Catarina

Downloads

Published

2026-10-29

Issue

Vol. 26 No. 1 (2026)

Section

Artigos

License

Copyright (c) 2026 Brazilian Journal of Business Economics

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

The submission of a manuscript authorizes its publication by the Brazilian Journal of Business Economics/Revista Brasileira de Economia de Empresas and implies that it refers to original work, which is not currently submitted to any other journal. It is forbidden any reproduction, partial or total, by any other media, either press or digital, of the published papers. In this case, previous and necessary authorization must be requested. In case it is obtained, one must include a formal acknowledgement to the BJBE/RBEE, which is the first publisher of the paper. The BJBE/RBEE does not pay any copyright to the authors of published papers.

Copyright and Licensing

The copyrighter of the BJBE/RBEE is the Editora Universa of the Universidade Católica de Brasília.

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Submitting an article authorizes its publication and implies a commitment that the same material will not be submitted to another journal.

The journal does not pay royalties to the authors of published articles.