Increasing position diversity by removing too correlated Positions
Author: Carova
Creation Date: 6/14/2017 4:49 PM
profile picture

Carova

#1
I would like to decrease the correlations among my positions during a backtest. Short of rewriting the strategy as a rotation, is there any way to test and remove positions or avoid Buys that are "too correlated" with existing positions? Thanks!

Vince
profile picture

Eugene

#2
Vince, you seem to be familiar with Correlation. What presents difficulty?
profile picture

Carova

#3
How do I access positions in the other symbols from the current symbol in a non-rotation strategy?

Vince
profile picture

Eugene

#4
Only by rewriting it as a Strategy that loops by DataSetSymbols.

And although removing existing Positions would not be an option, you could test for too positively correlated Positions in a custom PosSizer and avoid corresponding entry signals.
profile picture

Carova

#5
Thanks Eugene! That is what I thought. I will re-write the strategy as a Rotation and compare it to your suggestion of the loop. It is just that on a Daily strategy any of those loops really slow down optimizations.

Vince