Index | Recent Threads | Who's Online | Search

Posts: 22    Pages: 3    Prev 1 2 3
Last Post
New Thread
This topic has been viewed 884 times and has 21 replies
primus
Re: beyond sectors: ways to classify industries

Thanks, Yuval.

That sounds like a great approach! I might dig into doing this for myself.

"The world is. The world is. Love and life are deep maybe as his eyes are wide." - Rush, "Tom Sawyer"
http://the-world-is.com/blog

Jan 30, 2019 6:01:55 PM       
primus
Re: beyond sectors: ways to classify industries

Three part question here:

1. How are price benchmarks within ETF classifications calculated in P123? I know how it is done with stocks (i.e., a modified capitalization weighting with some filters). But what does "Close(0,#Style)" in an ETF context mean?

2. Is there a way to create price benchmarks of ETFs based on multiple criteria? Let's say I wanted to create price index of ETF within the same Style and Asset Class. Could I do this natively in P123 going back in time?

3. Has anyone thought of or found a way to create performance benchmarks based on quantitatively similar securities within P123, going back in time, and by using native P123 syntax. This could be similar to Yuval's correlation clustering algorithm, but doesn't necessarily have to involve Pearson correlation.

The reason I ask these questions is that I think they are germaine to parallel discussions on global equity momentum. I strongly suspect that any remaining inefficiencies in trend following going forward depend largely on selecting/constructing a good benchmark against which to measure performance. The challenge, however, is that a good ETF benchmark is not easily defined by looking at the cross-sectional performance against any single ETF classifier.

I mean, it makes no sense to me to measure a standard ETF against an inverse leveraged ETF in the same asset class. However, the performance of peers in the same asset class should be considered.

Maybe there's a simple way to cut through this classification problem?

*** QUICK UPDATE ***

I think it might be elegant to implement more granular peer groups through a “double sort” option. For example, this could be implemeted if composite nodes in of ranking systems were able to be ranked within a scope. The logic might look as such (the ETFComposite node and scope method in the node are not currently interpreted):

<ETFComposite Name="Cross-sectional Momentum" Weight="100%" RankType="Higher" Scope="Method">
<ETFFormula Weight="20%" RankType="Higher" Name="Relative Strength vs Asset Class" Description="" Scope="Asset Class">
<Formula>Close(0)/EMA(251,0)</Formula>
</ETFFormula>
<ETFFormula Weight="20%" RankType="Higher" Name="Relative Strength vs Region" Description="" Scope="Region">
<Formula>Close(0)/EMA(251,0)</Formula>
</ETFFormula>
<ETFFormula Weight="20%" RankType="Higher" Name="Relative Strength vs Style" Description="" Scope="Style">
<Formula>Close(0)/EMA(251,0)</Formula>
</ETFFormula>
<ETFFormula Weight="20%" RankType="Higher" Name="Relative Strength vs Size" Description="" Scope="Size">
<Formula>Close(0)/EMA(251,0)</Formula>
</ETFFormula>
<ETFFormula Weight="20%" RankType="Higher" Name="Relative Strength vs Sector" Description="" Scope="ETF Sector">
<Formula>Close(0)/EMA(251,0)</Formula>
</ETFFormula>
</ETFComposite>


Does anyone else think this a good and elegant idea?

"The world is. The world is. Love and life are deep maybe as his eyes are wide." - Rush, "Tom Sawyer"
http://the-world-is.com/blog

Feb 2, 2019 5:25:46 PM       
Edit 2 times, last edit by primus at Feb 3, 2019 4:22:31 PM
Posts: 22    Pages: 3    Prev 1 2 3
 Last Post