Exploring AI in Fantasy Football: A Unique League Experience
As the football season kicks off, I proposed an exciting idea to my hacker friends: a Fantasy Football league fully powered by large language models (LLMs). No human decisions allowed—every choice is made by a model. Here’s how we set it up:
- Model Selection: Each manager picks a model for the season, utilizing prompt engineering and various tools.
- Performance Tracking: Weekly tokens and decisions are shared among league members.
- Model Punishment: Poor-performing models can be swapped after a random grand jury assesses the situation.
Key learnings from our experience:
- Speed Matters: Initial time limits of two minutes per draft pick proved vital, but we faced unexpected challenges.
- Tool Utilization Over Parameters: Success often hinged on model adaptability rather than sheer power.
- Agent Efficacy: Surprisingly, simpler agents outperformed complex setups.
Curious to see if AI can dominate Fantasy Football? Join us on this journey, check out our drafts, and catch the banter on YouTube!
👉 Watch here and share your thoughts!
