Outliers, in my experience are where the losses really come from in a well designed system. When the market behaves in an average way, it can be characterized in a normal distribution curve which is easy to trade. Leaving out "bad" companies, leads to data mining bias and unrealistic results.
However, If you want to be conservative, leave out the top x performers, and leave in all the bad ones. ;-) BR, Dennis On Jul 3, 2008, at 1:31 PM, tipequity wrote: > any takers? Howard B? > > --- In [email protected], "tipequity" <[EMAIL PROTECTED]> wrote: >> >> Has anyone written code to exclude the top 5 winners and losers from >> the backtest and optimization results? It seems to me like a good >> practice to exclude x number of outliers from the backtest and >> optimization results! What do you think? Should this be something > that >> should be included in the backtest settings? TIA >>
