大多数预测市场的用户并不是在交易——他们是在赌博。
预测市场最近激增了700%,但从结构上看,它们更倾向于有信息优势的参与者。<p>83%的参与者处于净亏损状态,因为受情绪驱动的交易缺乏优势。学术研究表明,在金融领域,大型语言模型(LLMs)的表现优于人类预测者。能够对事件进行评分、监测波动性,并根据证据而非直觉执行交易的自主系统,可能是这些市场一直缺失的优化层。
查看原文
Prediction markets surged 700% recently, but structurally they favor informed actors.<p>83% of participants are net negative because sentiment-driven trades lack edge. Academic studies show LLMs outperform human forecasters in financial domains. Autonomous systems that score events, monitor volatility, and execute based on evidence—not intuition—represent the likely optimization layer these markets have been missing