Dec

21

One of the things I find important is stability in simulations. If your model exhibits one of the following instabilities, your chance of making money from them is smaller than you imagine!

  1. Instability with regard to simulation starting parameters. This means that the method is stable with regard to the choice of the start and end bar, small variations in sampling windows, etc. Examples of this instabilities: a. A method that makes money on a 252 day window and loses on a 251 window b. A classic example: the code from the book by Didier Sornette. He fits the market to a nonlinear exponential function using R except he doesn’t try refining and randomizing the starting points for the fit and changing the window. The outcome is a graph that “predicts” the market
  2. Stability issues with regard to small variations in the money management methods, say 10% position size, makes $$$$ and 10.04% loses. This brings the stability of optimal F parameters into play.
  3. Stability issues with regard to optimization - How much is done with regard to over-optimizing?
  4. Stability issues with regard to the choice of stock (I.E. in the case of equities)- say the model works on a certain class of stocks - it may not work on others!

These stabilities seem elusive. I am wondering if you have input and thoughts about these and other instabilities one


Comments

Name

Email

Website

Speak your mind

Comments are closed.

Archives

Resources & Links

Search