Automated Trader – 07-29-2008
Over a month has passed since I last checked the exact numbers on the trader. Maybe that’s because during the down period it basically sat on the sidelines and didn’t open a single real trade over a two week period. Maybe if I had allowed it to trade on margin, and thus short during the incredible downturn, I would have real profits rather than paper profits, but I can wait. So I’ve had 13 trades since mid-June and averaged 0.3 times the risk amount in profit per trade. It was a little more realistic this time with 46% of the trades ending positive after commissions.
The one thing I’ve learned is that the actual machine learning techniques you use don’t really seem to matter much. Whether it’s the simplest HMM, LDA, or SVM, the choice only shifts the results small amounts. Changing the inputs is what really gives you bigger gains. Of course it’s obvious that garbage in equals garbage out, but I’ve had to be cut down by Occam’s razor more than a few times when I’ve gotten overly complicated with too many fancy indicators and patterns. I guess the moral of the story is just to keep things simple and understandable and you’ll save yourself a lot of headache.
Comments
Leave a Reply