CHAPTER 10 Testing—The Fork in the Road

We’ve talked about testing already, but in generalities. It’s time for some clarity.

Testing is a necessary evil, so it’s best if you know the good and the bad of it. It is the development stage that can make you a success or turn a good idea into a failure. Only systematic traders can test properly because everything has been set down in clear rules. Putting your idea into a spreadsheet or programming it into a test platform, such as TradeStation, is a good way to audit yourself. You never realize what you’ve left out until it’s written down completely and you try following your own rules. Programming those rules is a sure way of doing that. Programming it in two different platforms, and comparing the results, is even a better way of avoiding errors.

There is as much discipline needed to control your testing as there is in trading. Computers are powerful and can just as easily give you the wrong answers as the right ones. They simply respond to what you ask. It’s very tempting to let the computer do more work and you do less, but that’s a mistake. I’ll try to show you the right way.

Start with a few basic principles:

  • More data is better. There is no such thing as bad data. The more data, the more market situations, bull and bear trends, price shocks, and sideways periods. They all need to be part of your test.
  • Leave some data out for validation at the end. It’s your reality check. Some analysts leave out the most recent data. I prefer ...

Get A Guide to Creating A Successful Algorithmic Trading Strategy now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.