Option FanaticOptions, stock, futures, and system trading, backtesting, money management, and much more!

Short Premium Research Dissection (Part 30)

I left off feeling like our author was haphazardly tossing out ideas and cobbling together statistics to present whatever first impressions were coming to mind.

The whole sub-section reminds me of something I read from Mark Hulbert in an interview for the August 2018 AAII Journal:

     > …people’s well-honed instincts, which detect outrageous
     > advertising in almost every other aspect of life, somehow
     > get suspended when it comes to money. If a used car salesman
     > came up to somebody and said, “Here’s a car that’s only been
     > driven to church on Sundays by a grandmother,” you’d laugh.
     > The functional equivalent of that is being told that all the
     > time in the investment arena, and [responding] “Where do I
     > sign up?” The prospect of making money is so alluring that
     > investors are willing to suspend all… rational faculties.

As discussed in this second-to-last paragraph, I miss peer review. What our author has presented in this report would have never made the cut into a to a peer-reviewed science journal. I think she has the capability to do extensive and rigorous backtesting and analysis. I just don’t think she has the know-how for what it takes to develop trading systems in a valid way.

To me, system development begins with determination of the performance measure(s) (e.g. CAGR, MDD, CAGR/MDD, PF). Identify parameters to be tested. Define descriptive and inferential statistics to be consistently applied. Next, backtest each parameter over a range and look for a region of solid performance (see second-to-last paragraph here). Check the methodology and conclusions for data-mining and curve-fitting bias. Look for hindsight bias and future leaks (see footnote).

System development should not involve whimsical, post-hoc generation of multiple ideas or inconsistent analysis. Statistics dictates that doing enough comparisons will turn up significant differences by chance alone. We want more than fluke/chance occurrence. We want to find real differences suggestive of patterns that may repeat in the future. We need not explain these patterns: surviving the rigorous development process should be sufficient.

Despite all I have said here, the ultimate goal of system development is to give the trader enough confidence to stick with a system through the lean times when drawdowns are in effect. Relating back to my final point in the last post, a logical explanation of results sometimes gives traders that confidence to implement a strategy. I think this is dangerous because recategorization of top performers tends to occur without rhyme or reason (i.e. mean reversion).

As suggested in this footnote, I have very little confidence in what I have seen in this report. On a positive note, I do think the critique boils down into a few recurring themes.

In the world of finance, it’s not hard to make things look deceptively meaningful. The critique I have posted in this blog mini-series is applicable to much of what I have encountered in financial planning and investment management. In fact, for some [lay]people, the mere viewing of a graph or table puts the brain in learning mode while completely circumventing critical analysis. Whether intentional or automatic, no data ever deserves being treated as absolute.