The best strategy I found was a simple Bollinger Band - purely because the market is so sentiment-driven (a nice way of saying "bubbly") that jumping on a bandwagon as it sets off was actually worthwhile, whether buying or selling. The sad fact is though, that once so many traders (and HFT bots) jumped in it's impossible to run a regular algorithmic trading bot. When the market starts to go into freefall then the bot is trading blind too. Frustrating but there you have it.

I've also learned a lot and have a nice sentiment analysis tool sat there in case of future requirements. Plenty of plans too. I haven't got around to my planned next project of running some sort of machine learning algo on historic data to optimise parameters for the Bollinger Band strategy. I still have a feeling Twitter sentiment could figure in that too but who knows?

Are you storing the twitter feed for further analysis? Running any textual sentiment analysis or just tweet volume?