Outcast Beta Blog

[jfan]

Home - Outcast Beta

 

This is a interesting blog that Corey Hoffstein supported on his podcast. He interviewed the author. 

 

I read through some of his articles and there are some interesting insights on portfolio management. 

 

He discusses the interaction between portfolio concentration, investment styles, & time and their effect on geometric outcomes and minimum threshold of required alpha.

 

Long-story short:

1) more concentration portfolios (<=5 (concentrated), 6-10, <20, >20) requires more alpha and increases risk of portfolio drag

2) probability of loss is less as time goes by but the magnitude of loss increases

3) probability of loss is less than expected < 5 years, but the magnitude of loss is greater than expected in < 5 year time frame according to real world data

4) microcap, high valuation, non-profitable companies, low momentum require much more alpha whereas as large cap, low valuation (E/B, E/P), profitable, high momentum require much less alpha

 

So Buffett's punch card of 20 stocks of profitable, value strategy is quite rationale, but with his concentrated approach, it makes alot of sense to keep lots of cash for opportunities. 

 

For an investor that likes to be fully invested all the time, it would be rational to take more of a diversified approach > 15-20 stocks and consider diversifying across styles to avoid any portfolio drag when a particular strategy is not in fashion. 

 

The author also mentions alpha consistency is a hard to measure variable but will contribute to portfolio drag, hence focusing on a repeatable process to optimize investment decisions tilts the table in your favor over the long-term.

 

I'm not a quant, but I did find the articles helpful in quantifying what I felt over time were key decisions in portfolio construction. 

 

The web address to the blog is above, and I've attached the author's thesis for those braver than I to dissect the mathematics behind it. 

 

Portfolio concentration, style, and alpha needed (thesis).pdf