Hello Don, yes this site can be used as a forum to influence AI improvements, but I also hope it will be much more.
I'd like to see some really cutting edge ideas and theories here. Even if none of them end up in AI, I think we might learn something we never thought about. And this germ of an idea, when combined with other germs, might lead to an idea that will end up in AI and improve everyone's returns as a result.
My view on multiple stocks in one portfolio is that there's no need to do it. I think you'd be better off with one stock per portfolio. The multi-stock approach, in effect, creates a new stock (or an index or whatever you want to call it) with its own unique characteristics.
However it becomes increasingly difficult to optimize this index because you've introduced new variables (i.e. each stock in the portfolio contributes something to the behaviour of the index).
The interactions between these stocks increase exponentially as each new stock is added to the index. The search space is already large for one stock (in a portfolio). If you exponentially increase this, it becomes clear that the time necessary to find the optimium settings will be too great (even with a time-saving tweak such as a genetic algorithm).
Optimizing one stock with the 6 standard optimization parameters set to their maximum ranges is already too much for a brute-force search. (See the AI 2.0 User's Guide for an example.)
In addition, I think that an index's past performance will not as accurately reflect its future performance as much as a single stock's past performance will reflect its future performance.
I do think that your suggestion would be a good one, but more as a research tool to indirectly gather insights, rather than a direct way to optimize multiple stocks in one portfolio.
Unfortunately there are a number of "need to add" enhancements, lined up like planes on an Atlanta runway during a thunder storm, that I have to address before I can even consider putting this one in.
Jack Park (http://www.thinkalong.com ) did some interesting stuff many years ago that you might find interesting.