I open-sourced my UFC prediction model, code, and database after 5 years of work (mcinerney.ai)

After five years of development and over 15,000 hours of work, Dan McInerney has open-sourced his machine learning model for predicting UFC fight outcomes. The model, which has achieved an impressive 8% return on investment (ROI) since 2024, is believed to be the first completely open-sourced Vegas-beating model in this domain. McInerney emphasizes that his project, hosted on GitHub and Hugging Face, includes not just the model and its weights but also a comprehensive database containing granular historical odds—offering an extensive resource for data-driven sports prediction. This endeavor is significant for the AI/ML community as it democratizes access to sophisticated prediction technology, enabling enthusiasts and researchers alike to explore and improve upon his work. McInerney shares insights from his personal journey, highlighting common pitfalls in machine learning practices, such as overfitting and data leakage, that he encountered along the way. The model is encapsulated in a Docker container with a user-friendly interface, allowing even non-technical users to engage with sports predictions. By cultivating a collaborative spirit and sharing his learning experiences, McInerney aims to foster further exploration within the niche of machine learning applications in sports.