In the latest Journal of Finance, there's an article by Ni, Pan and Poteshman about volatility trading. Basically, they argue that increases in demand for options by non-market makers helps predict realized volatility. OK, fair enough. Someone hears a rumor, buys options to capitalize, the rumor is realized, and volatility increases. Implied volatility and option volume help predict realized vol.
But here's where I lose interest. They use daily data from 1990 to 2001. The option market has become so much more liquid and efficient since 2001, I doubt anything in this period is relevant. Prior to 2001, it was pretty hard to trade this stuff algorithmically, picking up on things, because spreads were very wide (and still are), and quotes posted weren't very deep (ie, you couldn't do much size at the bid or ask), and systems for trading these things with electronic algorithms were very difficult in those days. Heck, even Nassim Taleb's fund made a lot of money in 2000.
There are lots of strategies that made money in the 1990's, such as any short-term mean-reverting strategy in stocks. Most of these are merely of historical interest, because they are long gone. High frequency data from 10 years ago is about as interesting as reading that if you created a search algorithm for the interweb (aka World Wide Web), you would be rich! It's true. And sharing videos online, another money maker.