Friday, October 23, 2009

PD FXY Update

I'm back . . with a corrected N day algorithm thanks to Jeff's prompt turnaround. The current signal is actually not LONG but net OUT since the Firing Line filter (still on progress) would override the FXA and FXF long signals short term and the FXE long signal longer term.
Tomorrow we'll line up the momentum swings in the FXY relative to the trade reports of both the short term and long term models. I've been holding off this task until the Dashboard was in a more stable condition but, with the latest changes to the database, I believe we're there.
Stay tuned. . .this could be interesting.



4 comments:

bill said...

bzb,

did IWM happy trail triggered a buy on 23rd.

Also i would appreciate if you can do a post on pair trading where you back test it for 6 months in 2002 to 2008 and then forward test and see if it works.

as always your help is appreciated and respected.

Bill

bzbtrader said...

Bill,
The 2002 -2008 thing isn't going to happen. As I keep mentioning, these pairs are transitory and you discard them and find new pairs when the equity curve fails to keep pace with the R2 slope. That's the basic idea behind the basket approach and the PDQ Dashboard. I wouldn't expect 2002 pairs to look like 2008 pairs ... the market dynamics, volatility, etc are entirely different. I'm not one who says that a system has to perform in a linear fashion for 5 or 10 years to be useful. Things change and I'm much more likely to adapt to those changes than try to force a square peg in a round hole.

stuti said...

bzb,

What i was thinking was about trying to research and see if we can test say between 2002 july to dec and say trade it for 2003 jan-feb-march, so need to do roll forward this kind of testing and see how it works.

Also did your Happy trail trigger.
I am running amibroker code and so lagging behind a bit



thanks
Bill

bzbtrader said...

Bill,
With original settings, HT on IWM is currently out and flat from a long position as of 10/14.
Re the 2002 thing....we can't pull data back from that far (yet) and that would involve a bunch of work to construct a basket of releavnt FXY ETFs hot on that time span, which, although interesting to think about, does not fit within my very overextended work load.