No. of Recommendations: 15
* The backtest of each half of the switch is cut approx in half so is less robust.
* You can tune by picking various indices and screens to switch on.
* You can tune the cut-off

All of this really means that the burden of stress testing the backtest is so much greater - i.e. it is that much more difficult to figure out if the backtest has predictive value.

Well, my thoughts were that you wanted to avoid much in the way of
backtest tuning at all. It's pretty easy to pick some blends just by
reputation: you're going to want things like RS26 and Screamers
in the bull periods, and YldEarnYear and ZLTDA in the bear periods.
It's pretty obvious once a bear is well under way (say, January
this year) that it's time to quit the momentum screens and think
about the dividend yield screens. So, if it's common sense, then
it can probably be usefully approximated in real life.

Doing it mechanically isn't hard, but it's circular: I'm using the
backtest results to pick what to use, then seeing how it works in
the backtest. Not so great for predictive value!

But taht never stopped me. I ranked all the VL screens by CAGR in bull
periods, CAGR in bear periods, and rolling-two-month downside deviation
calculated across all periods.

I figured rank(downdev)-rank(cagr-bull) would be a good way to find hot
stuff for bull markets with reasonable risk containment, and then use
rank(downdev)-rank(cagr-bear) for the bear periods.
In essence, I used a good rolling-2-month downside deviation across the
whole history as a proxy for a "good enough" screen. If it tends
not to lose money in two-month periods often, it's probably an OK
place for your money. (why 2 months? it reduces noise inherent in
only a few picks at a time, and eliminates problems the fact that a
lot of screens have positive or negative month-to-month correlations).

The funny thing is, the blend that this picks for the bear periods
works so well in the bull periods that you might as well use it
all the time and not bother switching!

Here's what it came up with.
For bear periods
For bull periods

The bear screens backtest at around 35.2% overall, the bull screens
at about 37.8% overall, and the switch at around 41.1%.
That's using a simple "S&P above 12 month-end SMA" test.
For this purpose, the optimal is "above 6 month SMA", which ups it to 43%.
But, the bear blend is a steadier earner than the switch, and looking
at the equity curves, it's a much better and simpler choice!

Here are the bear blend returns.
1989   62.5
1990 -1.2
1991 76.7
1992 45.7
1993 45.2
1994 13.4
1995 55.3
1996 24.6
1997 48.7
1998 22.9
1999 43.7
2000 39.0
2001 101.4
2002 24.9
2003 40.3
2004 43.1
2005 44.6
2006 27.3
2007 33.8
2008 -2.9 (that's Q1 only--TTM is +22%)

Executive summary (one possible interpretation): if you can pick a
long-only blend that works in bear periods, it's probably an excellent
blend for bull periods too, so why bother with the switching?

Print the post  


When Life Gives You Lemons
We all have had hardships and made poor decisions. The important thing is how we respond and grow. Read the story of a Fool who started from nothing, and looks to gain everything.
Contact Us
Contact Customer Service and other Fool departments here.
Work for Fools?
Winner of the Washingtonian great places to work, and Glassdoor #1 Company to Work For 2015! Have access to all of TMF's online and email products for FREE, and be paid for your contributions to TMF! Click the link and start your Fool career.