Originally posted by wilfran1pm
No problems Thanky. It sounds lie you are on the right track. Handy's stuff is a bit challenging to get through. He isn't what I'd call a dynamic presenter, and his books are more of a tutorial rather than a descriptive easy read. He does breach the subject of machine learning as well. I haven't found that much info on application of MC to trading systems at all. It is something that interests me and I did complete the Stanford University Open learning on-line MC course. I had to do a quick refresher course on Linear Algebra and learn how to use Octave, the freeware version of Matlab, but it was all worthwhile. I've listened to some podcast by people who apparently use MC in Trading but none of it disclosed exactly how they use it or how they build their models. They all just come out with wild statements about how surprising and wonderful the results are. No mention of all the work involved in building the models. MC relies on what turn out to be very very large two dimensional matrices that consist of vectors of what are referred to a "features" which are associated with a known outcome. A feature could be the one period ROC of a 5 period Stochastic for example, or even the 3 or more period shape of the Stochastic curve, and the outcome would be the resulting move in the SP. The number of features you could include is almost infinite. So you take lots of samples of many features and the associated outcome and then run one of the many MC algorithms over the data to learn exactly which combinations provide the best outcome. This is all fine but you still can't avoid all the work involved in determining the features and then compiling the feature and associated outcome data. I think the best approach would be to develop some sort of indicator based model that seems to make sense for the problem you are focusing on and then move forward with MC to refine the settings and determine the indicator movement and positions that provide the optimum outcome.
Not all that easy I think. And if there was ever a rabbit hole trap in something I think it may be in the application of MC to stock trading.
If you find any useful info on MC and its application to trading, let me know. It's so
mething I'd like to follow up on sometime in the future.
Hey wilfran and
@Thanky this guys podcast on machine learning in systems is pretty good. He also has a blog/website that has more info on it too
http://bettersystemtrader.com/082-machine-learning-with-kris-longmore/
thanky I started down the same rabbit hole about 12 months ago it’s actually pretty fun but all consuming
(20min delay)
