Pardo - Measuring Degrees of Freedom

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.
User avatar
strix
Roundtable Fellow
Roundtable Fellow
Posts: 65
Joined: Mon May 24, 2010 2:48 am

Pardo - Measuring Degrees of Freedom

Postby strix » Wed Aug 18, 2010 9:47 pm

Here are a couple questions for any of you guys who have read Robert Pardo's book, "The Evaluation and Optimization of Trading Strategies"...

1) When measuring the degrees of freedom in a CSI Futures data sample, would it be correct to compute this as

6(OHLCVOI)*Number of Years * 260 * Number of Markets ?


2) On p.292 he talks about a dual moving average system and says each moving average uses as many degrees of freedom as its length, plus another degree of freedom "as a rule".

How do you count the number of rules in your system? If you have, say, a long entry signal that uses 2 different moving averages, is that one rule or two? Once you start combining indicators, it seems like things get tricky here.

Any advice would be appreciated. It's not the clearest book in the world.

User avatar
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2973
Joined: Fri Jun 11, 2004 2:50 pm

Postby sluggo » Wed Aug 18, 2010 10:19 pm

If you are having trouble interpreting the second edition of Pardo's book, perhaps you might want to contact the author, Robert E. Pardo. Nobody understands the subtleties and nuances of a book better than its author.

Mr. Pardo runs an investment advisory firm named Pardo Capital Limited which has a website: http://www.pardocapital.com/contact.php5 . According to the website you can send email to Robert E. Pardo himself using this email address
  • rep atsign pardocapital dot com

Good luck and best wishes on your project!

I don't personally know what Pardo means by "degrees of freedom" but I think it might be sensible to distinguish between (things that matter) and (things that are irrelevant). For example, Trading Blox software ships with eleven presupplied trading systems, and none of them use Volume or Open Interest when calculating entries and exits. So, at least for this particular set of eleven systems, Volume and Open Interest are irrelevant. Now whether that means V & OI are, or are not, "degrees of freedom" according to the official Robert Pardo definition, I do not know. However since V & OI are irrelevant, at least for these eleven systems, I don't quite see why it might be useful to count them.

For a second example, some of the eleven presupplied Blox trading systems, use ONLY the "Close" price when calculating their entry and exit signals. These systems pay no attention to the Open or to the High or to the Low. For these systems, Open, High, and Low are irrelevant. But would Pardo count Open and High and Low as "degrees of freedom" nevertheless? Again I don't know. Is it useful to enumerate irrelevant data? I have my doubts.


Return to “Testing and Simulation”

Who is online

Users browsing this forum: Yahoo [Bot] and 2 guests