gbarber Veteran   Posts: 282 Joined: 12/30/2012 Location: Pearland, TX  User Profile | Geoff The last paragraph you wrote reminded me of a previous post by you. It got almost no notice but I believe deserves considerable attention. I insert the link below for those who have not read it. http://www.omnitrader.com/currentclients/omnivestforum/thread-view.asp?threadid=7354 What you propose in that thread seems to me to at least the beginning of a great method for determining the predictive reliability of a portfolio. One could define a boatload of statistics on the results from runs walking across a broad time period. The stats would include things like variance, standard deviation and the like. It seems to me that the lower those values are, the more predictive the portfolio is. That is the CAR and MDD should hold at a pretty flat value. If they do one can believe that the portfolio is predictive. The stats would be a measure of the confidence one can have of the predictability. For those who have not read Geoff's post, especially Nirvana folks, please do. It seems to me to be a great idea. | |
