Thursday, August 13, 2009

Equity Curve Surfing

Here is one last look at the unbounded DV-2 strategy by CSS Analytics to demonstrate the larger concept of Equity Curve Switching. From the last post (here), I took the two best equity curve results, namely those from the Raw DV-2 (blue) and the Opposite-Direction (purple) strategies, and then overlaid a simple switching routine based on the higher short-term rate of change among the two.

The result is the orange equity curve (simple, frictionless returns). Obviously, this type of system-level trading overlay will work best when the alternative curves present persistent offsetting periods of over/under performance.

In that last follow-on DV-2 post, I referenced the importance of "knowing your trading environment." Now imagine a similar system to the one described above that perennially checks and compares the performance of basic trending, break-out and mean-reverting strategies. In this regard, a good recent companion thought piece to this post may be found over at MarketSci on the State of Mean-Reversion Systems.

I've about beat the unbounded DV-2 indicator to death at this point (with respect David)! In the future I will work with the bounded variant.

Note: This post was originally published as an ETF Rewind 'subscriber-first' article delivered over the weekend. This will be my new practice to provide added value to my members.


Woodshedder said...

Jeff, is it possible that you could specify your ROC measure that determined when you switched?

"the higher short-term rate of change among the two."

I love this trading the equity curve series Jeff. Well done.

Jeff Pietsch CFA said...

Thanks Wood -- In this case I used a simple Last Price/ Average(N Periods). The strategy worked across a fairly broad short set of N's. However, the main idea is to think about applying relative strength methods to your equity curves in general rather than specifying a specific formula here.

Shivar said...

Jeff, I have tested your idea with the daily follow through concept. I wanted to switch between mean reversion and trend following so I build the equity curves of "going long if S&P500 daily perf > 0" and "going short if S&P500 daily perf > 0".

Then I defined your short term rate of change and switched to the strategy with the higher rate of change. The weird thing is that it fails pitifully if I am using a short term time window (like 20 days). The combined equity curve is worse than the worst of the initial equity curves.

If I increase the time window, performance increases (almost linearly).

Can you precise your concept to allow short term rebalancing ?

Jeff Pietsch CFA said...

Hi Shivar, I suspect you are headed in the right direction, but may consider reconfiguring your trend-method formulation. Think of trend-systems you would actually trade, and start with that. Best, Jeff

Dennis said...


I was wondering if you tried this approach as well with more more systems. Like 100 or so...


jgpietsch said...

Hi Dennis, I use it for about a dozen at a time. More on this in future posts. Best, Jeff