Importing strategy preferred values into Excel
Author: superticker
Creation Date: 1/24/2017 7:06 PM
profile picture

superticker

#1
Is there a tool that would allow me to import all the saved preferred values in a WL strategy into Excel? The tool would read the <PreferredValues> tag of a strategy XML file, then write those values to Excel.

My goal here is to have Excel return the mode (most frequent value) of each established preferred value for each stock. I realized the Results tab of the Optimizer tab already computes their arithmetic mean, but the mean always places the preferred value somewhere in the center of the range; whereas, the mode may place it at one of the extreme ends of the range (which is what I want).

For optimizer execution, it's best to set the default preferred value to the mode (most likely value) rather than some center-of-the-range value (i.e. the mean).

I may also want Excel to produce a histogram of each preferred value distribution to visualize how often these values approach the extreme ends of their range.
profile picture

LenMoz

#2
QUOTE:
For optimizer execution, it's best to set the default preferred value to the mode (most likely value) rather than some center-of-the-range value (i.e. the mean).
Not true for the Particle Swarm Optimizer, which uses the Hammersley algorithm to assign initial parameter values. The objective of the Hammersley algorithm is to produce particles that uniformly span the solution space. Default preferred value has no role in PSO.
profile picture

superticker

#3
QUOTE:
The objective of the Hammersley algorithm [for the Particle Swarm Optimizer] is to produce particles that uniformly span the solution space.
Thanks for pointing that out. I love the Particle Swarm Optimizer, and use it almost exclusively. It's very fast, and it gets the parameters values close enough for a fuzzy system optimization.

I think the Generic Optimizer is too precise for fuzzy system parameter solving. We really want the optimizer to return "close" in-the-ball-park values, which are time independent. For time-dependent varying values, it's the adaptive indicators' job to zero in on those with some precision.

I think the Wealth-Lab User's Guide should emphasize that PVs are for time independent parameters; whereas, adaptive indicators are for time dependent parameters. Not all WL users are numerical analysts and appreciate this distinction immediately.