Now that we have clearly established the limitations and benefits of various data analysis processes, we can examine the distinct levels of data analysis. Although there may be other methodologies by which to delineate our data, I have found that generally there are three different levels to data analysis of trading systems: analysis by asset classes, year-by-year analysis of in-sample data, and analysis of out-of-sample data.
For examples of analysis of data by asset classes, I refer the reader back to Tables 3.2 to 3.10. Ideally, system developers would like to see smooth and evenly distributed profits throughout all assets within these tables. However, it is more important that a system displays solid performance vis-à-vis risk than that such performance is evenly distributed throughout all assets in our backtested data history. With this caveat in mind, let us compare the various trading system results shown in Tables 3.2 to 3.10 and attempt to draw some conclusions regarding the data.
For now, we will narrow the field of study to those systems that generated a profit to maximum drawdown ratio of 3.0 or higher for the entire portfolio (shown here as Tables 7.21 to 7.25). Narrowing the field of study ensures that we do not waste our time and energy analyzing marginally performing trading systems, which we have no intention of trading in real time.
Table 7.21 not only generated the largest net profits, but it also ...