[hft_throwaway]

A simple signal would be more shares bid than offered at the inside market. Imagine a market with 1000000 shares bid at $4 and 100 shares offered at $4.01, it's more likely to tick up than down in the very near-term. This isn't tradable though since the predictions aren't large enough to overcome the spread and you can't expect to join the 1000000 share bid and trade when it's good (the 1000000 shares that made you think it was going to move up all have to trade or cancel for you to trade, not looking so great anymore).

Real HFTs use other features of the order book, movements in related products (maybe a move in oil futures in the last minute impacts the price of an airline stock in the next 10 seconds, etc.), and so on. The signals are not really that complicated, but competitors eventually converge on knowing the same signals and compete them out of existence/profitability. Also, how you execute around your prediction matters just as much if not more.

http://queue.acm.org/detail.cfm?id=2534976 and http://queue.acm.org/detail.cfm?id=2536492 have some more examples. http://www.decal.org/file/2945 also has some ideas for old signals that probably won't work anymore.