Sorry sir, but your breakout blog post gave me the impression that you didn't really deal with breakout rules before. I was wrong haha. Also there is some intro to machine learning in the MAN AHL videos.
Well the big fat NDA I signed with Man means I can only talk about what they're doing once it becomes public domain, hence the confusion. GAT
For anyone who's interested, I have open sourced my trading system: Quick features: Downloading and merging of data from Quandl & IB Bootstrapping of rules/portfolio Uses multiprocessing and vectorization where possible, so it's quick. Integration with Interactive Brokers (managed accounts) for automated trading https://github.com/chrism2671/PyTrendFollow
I got it running, it's full auto, uses market orders using IB's adaptive algo. Not sure what the adaptive algo does though...
I used to, but there are too many things that can go wrong with data feeds etc. Normally I: Download the data with cron, as this takes a while (download.py). Run validate.py and inspect the output, which shows the status of price data for each instrument and how old it is. Run scheduler.py --now to execute trades.