R-Cubed

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
LeviF
Roundtable Knight
Roundtable Knight
Posts: 1428
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

R-Cubed

Post by LeviF » Mon Aug 17, 2009 11:06 pm

Has anyone else noticed the time component of R-Cubed makes it sort of a finicky objective function? I'm thinking RAR / avg max DD might be a more robust measure. I think that the length of DDs is already implicitly included in this measure anyways. A system with longer DDs is going to have a lower RAR by nature.

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Mon Aug 17, 2009 11:43 pm

Yet more evidence supporting my theory that no two traders can agree on what a "good" trading system's measured characteristics might be. This is why there are so many different ways to measure goodness (what Blox calls "Test statistics") -- because everybody has their own unique and idiosyncratic way to define and measure That-Which-Is-Good. Fortunately, software like Wealth-Lab and Blox provides a way for users to program their own personal favorites. Blox's "test statistics" are Wealth-Lab's "Perf Scripts", boo yah. You can program a measure of goodness, exactly & precisely YOUR WAY, and all other traders be damned. Books (such as "Way of the Turtle") be damned too.

I recommend you research "Peter Martin" and his "Ulcer Index" for additional inspiration in the art of unique and highly individualistic "what do *I* mean by GOOD" definitions. Then poke into "Lars Kestner" and his "K-Ratio". Bob Pardo and his "Percent Of Perfect Profit." Ed Seykota and his "Lake Ratio". Cynthia Kase and her K-Stats. Gosh there sure are a lot of these.

Asamat
Roundtable Knight
Roundtable Knight
Posts: 175
Joined: Fri Jun 03, 2005 7:50 am
Location: Walldorf, Germany

Post by Asamat » Tue Aug 18, 2009 4:30 am

As alway, Sluggo, you have a point. Still, I agree with levijean. R-cubed sometimes changes rather a lot, just because the length component changed a lot, while all other characteristic numbers are very similar. Therefore I now use both numbers, but I rely more on the ratio levijean mentioned. I call it rMAR.

corvus
Senior Member
Senior Member
Posts: 31
Joined: Sat Sep 01, 2007 11:41 pm

Post by corvus » Fri Oct 23, 2009 2:20 pm

Asamat wrote:As alway, Sluggo, you have a point. Still, I agree with levijean. R-cubed sometimes changes rather a lot, just because the length component changed a lot, while all other characteristic numbers are very similar. Therefore I now use both numbers, but I rely more on the ratio levijean mentioned. I call it rMAR.
Why stop there? Why not Robust Sortino? :wink:

Asamat
Roundtable Knight
Roundtable Knight
Posts: 175
Joined: Fri Jun 03, 2005 7:50 am
Location: Walldorf, Germany

Post by Asamat » Mon Nov 02, 2009 4:12 pm

corvus wrote:Why stop there? Why not Robust Sortino?
What for?

I need some measures to compare systems and parameter settings. They have to be few in number, simple and robust.

I don't use most of the measures with the fancy names, where I have to look up the definition time and again.

Post Reply