Micheal Harris (PAL developer) suggests doing portfolio backtests on "comparable securities", which I assume means stocks for stock, forex for forex, etc. Has anyone tried to do that on nn-comparable securities? I did and the results are not as robust I suppose due to the different market dynamics (vol, drift, etc)