You built a classification model to predict whether the S&P 500 will go up or down tomorrow. Standard 5-fold cross validation shows 65% accuracy, but it consistently loses money in live trading. What is fundamentally wrong with your validation approach?

你做了一个预测标普500明天涨跌的分类模型。标准的 5-fold 交叉验证显示准确率有65%，但实盘一上线就开始持续亏钱。你的模型验证方法出了什么根本问题？