May

13

There is something like the keech cult in many academic papers about systems. Many of them don't work in the real world. The more they don't work in the real world, the more the academic papers with titles like "is momentum really momentum?" or "the disposition to ride winners too long—" or "investing with style" or "value and momentum everywhere" or "dissecting anomalies" or "251 years of price momentum" or "the world's longest back test" exist. There seems to be no awareness of the principle of ever changing cycles to explain why things like fama french discovered in 1992 with retrospective compustat data, don't work in practice. Similarly why momentum strategies will reach a peak before a year like 2008 when they lose 85 percentage points relative to neutral.

Ed Stewart writes: 

I believe the failure to account for changing cycles and how the profit incentive inevitably leads to a reduction or removal of a static profit opportunity to cost or worse has to do with the academic's need to create an aura of prestige and permanence in their work. You don't see academics studying supposed profit opportunities in retail or other mundane businesses that lack (what often is) the veneer of intellectualism that finance offers. Reducing things to this reality would kill the profession.

anonymous writes: 

I'm glad the Chair brought up again the subject of Ever-Changing Cycles, a.k.a. "regime switch". Where could one find literature about it, aside from Bacon?

There's an entire industry of publishing about "alpha-generating signals", but I can't find one reliable source of how to treat data in order to be aware of the regime shift as soon as possible.

It seems this is the "Great Arcanum" of speculating, and even the liberated Adepts are not allowed to mention it.

PS: We just suffer the effects of regime switch recently, in a very frustrating way: after 3 months of incredible hard work to put together a portfolio of futures strategies (in order to go "full f" with it), we noticed that something wasn't ok with the distribution of outcomes. We then realized that PBR (Petrobras) which is "the" political brazilian stock, was being traded very differently than the previous years, due to this election year of 2014. It seems that the whole communist party (which PT - Partido dos Trabalhadores really is) is buying the stock, pricing it higher with great urgency, in order to make up the immense dammage they (the government) did with the company. (They could make an oil company to go broke).

But, that's post mortem reasoning: our portfolio remain useles, since PBR (Petrobras) has a great deal of impact in the stock futures contract (our primary trading vehicle).

The question that arises is:

What length of testing should be used as metrics for regime switch awareness?

What is the testing one should do to put aside a strategy that is performing badly?

In other words, what would be the procedures a spec should be doing to prevent being caught in the changing cycles?

We would appreciate very much any guidance.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search