I agree also I guess you know this Berny, but I would argue that just seperating stocks into 2 groups isn't enough. Also need to split by time - eg in-sample could be 2000-2004. Out-of-sample 2005-2007...