Backtesting Strategies

I have mentioned many times that one set of technical indicators does not fill all, and I will remind you many more times.  The core reason is that each indicator performs a little different, which in turn may produce from slightly to vastly different results.  Even the same indicator with minor changes in parameters can produce widely different results.  To get you started, I have tried to point you in the direction of some simple indicators which you can understand, and which are most likely to produce favorable results based on past performance. 

I have been working with technical analysis for more than 25 years and have a top of the line charting and analysis package, both of which give me some insights into how particular indicators have performed in the past.  I can quickly pull up a chart with my favorite indicators and see if it is tracking a trend that I would use to make decisions.  But let there be no doubt.  It is a subjective process that leaves ample opportunity for misinterpretation and wrong conclusions. 

Ultimately, you need to see objective backtesting results.  Backtesting is terminology used to refer to testing a strategy or model using historic data.  Backtesting is an attempt to estimate the performance of a strategy if it had been used during previous periods.  Obviously, the strategy must have very specific rules which can be clearly stated so there is no ambiguity about the decisions that would have been made.  Clearly, the past performance of a strategy does not guarantee that it will perform well in the future, but it is a process which is effective at screening out strategies that did not work in the past.  One of my favorite sayings is that “if this horse has not won any previous races, it is unlikely it will win the next race.  Possible, but unlikely!” 

If you thought finding a charting and analysis package was difficult, try finding a package to do backtesting.  I know of some on trading platforms for stocks and bonds, but have been unable to find any that are reasonably priced to make it practical for beginners.  When I got relaxed on a recent fishing trip to Ontario, I had a brain storm.  There was no reason that I couldn’t extend the concept in the spreadsheet templates, which I posted earlier, to include backtesting of historic data.  

The most obvious limitation is the need for detailed historic data.  With years of historic data on my computer, that should not be a major task for me.  But it is.  I am working with my data vendor, but to date have not been able to get direct access to open the data files.  Through the assistance of a friend, I have been able to obtain five years of historic data (2008-2012) for old crop and new crop corn, soybeans, and wheat.  The extended template is evolving, and I have preliminary results for wheat that I will post in the near future. 

In addition to backtesting strategies which we have discussed, the new template makes it relatively easy to define a new strategy and get almost instant results to compare against the performance of other strategies.  Needless to say, this opens a whole new realm for those with minimum computer skills who are interested in learning more about how specific strategies perform.  The good news is that it can all be accomplished with very little cost other than time.  Nearly everyone has a spreadsheet program on their computer these days, and I have found a quote of $49.95 for fifteen historic contracts.  So, for fifty bucks and an investment of some time, you can get set up to do your own backtesting.

Comments or questions are welcome and encouraged by using the Contact Us page or responding directly to the post on the Discussion page with similar title. 

Posted by Keith D. Rogers on 28 July 2013.