Option-Driven Sentiment in FinRL: a PPO Approach to Trading

This paper addresses Task 1 of the FinRL Contest 2025 by applying a sentiment-enhanced stock trading agent based on Proximal Policy Optimization (PPO) augmented with market sentiment indicators, such as Put-Call Ratio, as key input, which is an option-based market sentiment indicator. The agent was tested on NASDAQ-100 constituents from 2019 to 2023 by using the trained dataset from 2013 to 2018. We found the PPO model outperforms both the baseline index and CVaR-PPO (CPPO) in cumulative return, Sharpe ratio, and Sortino ratio. Notably, the Sharpe ratio improved from 0.69 to 1.08. The results suggest that incorporating market sentiment indicators into PPO can significantly enhance risk-adjusted performance.