This is a follow-on from this post.
My sample size is now 251. I did some more tinkering and discovered that a time-based filter seemed more useful than a range-based filter. This is the equity curve for all trades taken between 00:00 and 14:00 GMT, which roughly cover the Asian, European and Europe/US overlap sessions. It seems to do poorly during the US session, which shouldn't be too surprising.
Reward-to-risk = 1.75:1. Risking 1% per trade.
Now this is starting to look interesting. Max drawdown is currently 11.9%.
Interesting. As always, in awe of your backtesting prowess.
ReplyDeleteIndicator for metatrader here if you don't have one.
https://drive.google.com/file/d/0B6FCA9dsxsq9U3hvWWxBNG1DT3c/edit?usp=sharing
One potential modification - only take up shaved bar long if the most recent shaved bar before it is a down shaved bar and let at least some of the position run until a down shaved bar forms (and vice versa for shorts). May increase the chance of catching large moves, avoiding chop, improving win rate and win: loss ratio. - at the expense of opportunity which can be at least in part offset by trading across multiple pairs and the potential to trade larger position sizes with a similar potential max peak to trough drawdown.
Hey thanks. :)
DeleteI've just about tested 2010, 2011, 2012 and first half of 2013. The system performed well until mid-2012. I'll post some more details in a future post. There was some interesting findings, especially on the usefulness of time-based filters. What you've described (going long after a down move, and vice versa) also seemed to work really well. I still got alot of data to crunch, though...