Filters

What is the general opinion on using filters in simple technical systems? Do they add value in general, or are they just another source of curve-fitting?
 
Anything which stops you entering trades is actually reducing your sample-size in back-testing, which will lead to curve-fitting. So to test if filters are working, use walk-forward testing.
 
Adding filters is just adding parameters to your model/trading method. So yes, you are curve fitting if you add multiple filters
 
Quote from FaceOff:

Anything which stops you entering trades is actually reducing your sample-size in back-testing, which will lead to curve-fitting. So to test if filters are working, use walk-forward testing.

Agree, to a point.

Switching a fairly simple trading system on or off based on it's equity curve can lead to a more robust result than adding additional filters directly to the system's trading rules.

Like anything else though, YMMV...
 
Quote from jprad:

Agree, to a point.

Switching a fairly simple trading system on or off based on it's equity curve can lead to a more robust result than adding additional filters directly to the system's trading rules.

Like anything else though, YMMV...

Without more specifics, my mileage is zero :-/

I assume you'd turn it on when the equity curve is showing a good ol' 45 degree climb and off when it hits an account balance trailing stop?
 
Quote from Dobbes:

What is the general opinion on using filters in simple technical systems? Do they add value in general, or are they just another source of curve-fitting?

If the performance is increased without decreasing the statistical significance of the results you may have something there.
 
Back
Top