Hello Everyone,
I was studying System Testing Chapter from Perry Kaufman’s book Trading Systems and Methods – 5th Edition and found the following text to be a bit confusing. Can anyone please break it down.
TEXT:
Alternating random periods are the most robust testing method, but those periods must be fixed at the beginning of the development process and not changed. If you were working on two different systems, each one would have its unique set of test data. In that way, you are not always testing the same data for every method you create. To create random test periods requires a small program, but that could be implemented on a spreadsheet.
• Decide how many in-sample periods and multiply by 2. Let’s say there are only 2 in-sample and 2 out-of-sample periods
• Does the in-sample come first? Let’s assume yes.
• Then, for 20 years of data, each segment will have about 5 years. Let’s say that the shortest will be 2 years and the longest 8 years.
• Start the first in-sample period on day 1.
• Find a uniform random number (between 0 and 1) and multiply that value by 6, the range of years in the in-sample period. If the random number is 0.72 then the adjusted value 4.32 is added to the 2-year minimum giving the length of the first period as 6.32 years.
• The same calculation is repeated for the next two segments. The last segment is whatever is left over.
Once calculated, these intervals can be assigned to specific data and all in-sample testing will use that data. This data can be used as much as needed. Rules and parameters
can be changed, and a detailed inspection of performance should be done. The data can be tortured without mercy.
Can anyone please breakdown the BOLD text in simple terms as it is a bit confusing for me.
Thank you.
Regards,
Maisam
.
