I'm a software engineer with a background in AI/ML, and this has been a side project of mine for a while. I was always fascinated by options trading but frustrated with the "get rich quick" noise online. I wanted to see if a systematic, data-driven approach using Reinforcement Learning could actually work.
I chose 0DTE SPX options because they create a perfect, self-contained "episode" for an RL agent to learn in—no overnight risk, cash-settled, and a clean slate every day.
This is the first post in a series about the project. It covers the setup, the strategy (Iron Condors), the model (Recurrent PPO with LSTM), and the surprisingly positive results from my initial simulation on OHLC data.
The key takeaway, however, is that this initial success is just one part of the story. The next posts will dive into the harsh reality of bid-ask spreads and why the model's performance cratered when faced with more realistic data.
Would love to get feedback or hear from anyone who's tackled similar problems.
