Editorial illustration for AI Agents in Charge of Fantasy Soccer Team, Not Better Than Humans Yet
AI Agents vs Humans: Fantasy Soccer Showdown
AI Agents in Charge of Fantasy Soccer Team, Not Better Than Humans Yet
My fantasy Premier League team has been managed by a committee of AI agents for several weeks now, and the results are humbling, mostly for the AI. The experiment started with a simple question: can a swarm of bots actually pick a winning fantasy soccer squad better than a person scrolling through injury news on their phone? This isn't the first attempt.
A few years back, a tool built by the UK's Turing Institute tried modeling past fantasy league data to suggest lineups, and the results were strange. It had no idea when a player got hurt or when a manager got sacked. It just crunched numbers in a vacuum.
This round is different. Using OpenAI's Codex, I built a harness, essentially a set of instructions and code, that pushes multiple agents to scan news, parse research papers, and generate predictions before settling on a team. The setup pulled in far more information than I could process on my own, checking sources and running models while I was busy doing other things, like sitting on a plane. The early output, though, still managed to make a call that any casual fan would question immediately.
Interestingly, AI still often lags behind very skilled experts known as superforecasters. Xu explains that, although humans cannot consume as much information, their understanding of the real world—that is, of the dynamics behind a news story—is deeper.
Why this matters
A fantasy soccer bot losing to a human picking players by gut feeling isn't a gotcha against AI, it's a reminder of where these systems actually stand right now. Xu's point about newer models pulling in more real-world data and coordinating multiple agents at once is the thing worth watching, not this particular season's standings. We've seen this pattern before with AI tools in trading, forecasting, and scheduling: the first version looks like a toy, then the data pipeline gets better and suddenly it's eating lunches.
For developers building agent swarms for anything prediction-heavy, injury reports, pricing, scheduling, the lesson here is that orchestration matters as much as the underlying model. A single strong LLM making picks in isolation is a different product than several agents dividing labor and checking each other's work. Don't read "not better than humans yet" as "won't be." Read it as a snapshot of a fast-moving baseline, taken at a moment when the hard part, coordinating multiple models on messy real-world inputs, is still being worked out in public.
Related Reading
Market Trends AI drives three initiatives to accelerate global action for planet protection
Market Trends Anthropic Burns Gas to Feed AI Boom Amid Intense Competition for Power
Market Trends NVIDIA and Google Cloud let developers scale AI from prototype to production
From Archives Anthropic sandbox-runtime OS-level isolates AI agents, denies network access
From Archives AMD Says Server CPUs Are a USD 200 Billion Market as It Announces Helion Blueprint
Common Questions Answered
How did the AI agents perform compared to human fantasy soccer managers in this experiment?
The AI agents managing a fantasy Premier League team underperformed compared to humans, with the results being described as humbling mostly for the AI. The experiment demonstrated that even with algorithmic optimization, AI has not yet surpassed skilled human players at fantasy soccer team management.
Why do superforecasters and skilled humans outperform AI agents in fantasy soccer predictions?
According to Xu in the article, although humans cannot process as much information as AI systems, their deeper understanding of real-world dynamics behind news stories gives them an advantage. This contextual knowledge of how events actually unfold in the real world allows skilled humans to make better predictions than current AI models.
What was the previous attempt to use AI for fantasy soccer predictions, and how did it perform?
A few years ago, the UK's Turing Institute built a tool that modeled past fantasy league data to suggest lineups, but the results were disappointing. This earlier experiment established a pattern showing that initial versions of AI tools in competitive domains like fantasy sports often underperform human expertise.
What improvements does the article suggest could make AI agents better at fantasy soccer in the future?
The article highlights that newer AI models pulling in more real-world data and coordinating multiple agents simultaneously show promise for future improvements. This pattern of evolution has been observed in other AI applications like trading, forecasting, and scheduling, where initial versions improved significantly as data pipelines and agent coordination advanced.
Further Reading
- I Put a Bunch of AI Agents in Charge of a Fantasy Soccer Team. They’re Not Better Than Humans—Yet - WIRED
- AI in Sports and AI Agents, the Game Behind the Game - PwC
- Toward Autonomous Artificial Intelligence Agents in Sports - PubMed Central
- Inside Bevo Lab, U. Texas’ Plans for “the Best Analytics Platform in All of Sports” - Google Cloud
