• Twitter可能预测过去,但很可疑能不能预测未来
  • 预测本身就是对未来的干扰,不可证伪
  • 过去的相关性不能代表未来的因果性



To be honest, I’m skeptical about the work in a couple of ways.

First, as some others already pointed out, correlation does not necessarily
mean causality.

Second, I’m not sure if it is more accurate (in predicting) than existing quantitative measurement of market calmness (or the lack of it) such VIX [1], or put/call ratio for an individual stock. While I don’t have any statistics, I personally find put/call ratio of GOOG is usually more informative (and direct) than reading tweets about $GOOG. Of course, their work is about the market as a whole, which I have no clue. Why their result is for 2-6 days later,
but not the next day?

read more