AI agents manage fantasy soccer team, but still fall short of human expertise

A researcher tested AI agents in managing a Premier League fantasy soccer team, finding they still lag behind human experts. The experiment, conducted in October 2026, involved deploying over 100 AI agents to analyze news, results, and player data. Despite improvements, the AI’s suggestions did not outperform the average human player. Haifeng Xu, a computer scientist at the University of Chicago, noted that newer models are better at processing real-world data and orchestrating complex systems. Xu’s team tested AI on prediction markets, finding models like GPT-5 and Grok 4 outperformed many humans. However, superforecasters still edge out AI in understanding contextual dynamics. Xu believes AI will eventually match human predictive abilities, but for now, the best human traders may remain in charge of AI-driven strategies. The experiment highlights AI’s growing role in prediction, though it has not yet surpassed skilled human forecasters. The study also references benchmarks like Prophet Arena and ForecastBench, which track AI’s predictive capabilities. While AI models consistently lost money in prediction market experiments, they showed steady improvement over time. The researcher concludes that AI’s predictive power could disrupt industries reliant on forecasting, but for now, human oversight remains crucial. The experiment underscores the ongoing evolution of AI in understanding and predicting complex real-world scenarios.