Fit for the Brexit? A new Smart Alpha Model released...

Hi all,

I’d like to start this post with a quote from another thread . In there I listed among others the following reason why the P123 community might not be interested in Smart Alpha models:

And this was Marc Gerstein’s reply:

With this thought in mind I made an effort to create a comprehensive documentation for my latest Smart Alpha model 7sisters Ultradefensive 10 stocks . The documentation is available on the smart alpha page and also in the attached pdf.

In particular I wanted to emphasize on the following two aspects:

  • curve fitting (or lack thereof): like many other designers I initially fell into the curve-fitting trap and saw some of my ports miserably fail out-of-sample. As you will see in the documentation, I used rolling tests, various universes and hedging instruments to demonstrate that the algorithm has a chance to be robust out-of-sample;
  • comparing to similar models: a potential subscriber should have an idea to what type of existing Smart Alpha model the strategy can be compared to. In my case I choose Keating’s Ultra Defensive Portfolio with comparable cautious approach and risk/reward profile. Furthermore, Keating’s SA model has the highest Sharpe ratio of all SA models and is backed by over 30 subscribers. Over the past two years the two models look very similar, and even in the short 3 month out-of-sample period my new SA model has the same 2% performance.

Please get back with some feedback and let me know if anything in the documentation is unclear.

Happy post-Brexit trading!

Florian


SmartAlpha_7sisters_Ultradefensive_10stocks.pdf (1.07 MB)

Hi Florian,

Since you didn’t qualify high Sharpe with other measures, I feel compelled to point out that my SlingCharts.com: Super Defensive S&P 1500, Hi-Performance (Founder’s Tier)
has the highest Sharpe ratio of all SA models.

Best,
Walter

Hi Walter,

You are absolutely correct. I only compared to models which have at least 2 years of OOS performance, hence yours didn’t come into the picture. I will add this fact to the description.

Just out of curiosity I ran my SA model with your SP1500 universe starting from your model’s launch date. Below is the result.

Best regards,
Florian


Strange. I get very different results for your UD model. Is there a problem with my book sim ?

No, there is nothing wrong with your book sim. The statistics I showed above are using an altered model of my algorithm which includes SP1500. The SA model is only using SP500.

Florian,

So how does your SA model perform over the full timeframe with the S&P 1500 universe? If it is improved as much as your post above, you may decide to add a new SA model. :smiley:

  1. I would like to see the performance without market timing. 2) would like to know what market timing is based on. thanks, Debbie

Also, the in/out hedging is probably contributing to the high turnover rate.

I’m happy you want to document, but look at the quote you picked up from me . . . noting especially the parenthetical. If you want to add simulation info, you can do it, but the documentation should be able to stand on its own and be persuasive even if there is no reference at all to simulation data. That means writing about why the industries in which your model invests can legitimately be considered defensive and what it is that makes your ranking system worth using. (No test results! What is it about the ideas behind the model.)

Suggestion: Write a 100% complete documentation with absolutely no reference to any test information. Only after you’re satisfied that it’s complete, then add test info at the end as an Appendix.