I have a system that employs a cascaded approach to stock trade selection.  That is, stocks are culled through a succession of filters, from the coarsest to the finest.  The virtue of this approach is that we can save CPU time by doing as little work as possible on individual securities until we have highly qualified them as suitable candidates.

An obvious qualification for stocks is price and volume minima.  So for example, I might want to create a dynamic universe of stocks whose price is >$9 and whose 50dVMA > 500,000, on any given bar.  A simple filter like this cuts the number of securities in the universe from ~8,500 names to ~1,750.  This immediately eliminates 80% of the subsequent processing I would have to do without this qualifier.   I would then proceed to work with the much smaller universe.

When I talked to your sales department prior to purchasing, the gentleman suggested I do this processing at the trade plan level.  Having now spent some time with the product, which is impressive in scope and conception, I see that the problem with selecting tradeables at the trade plan level means that we will have already had to run the rest of our work without regard to a security's satisfying the high pass price and volume filters.  In fact, having played with your NDX system and set up a single variable, 3-interval optimization and knowing the overhead associated with my back end work, I believe that salesman's proposed solution will be prohibitively slow.

So here's my question: what is the best way to create dynamic stock universes on a bar-by-bar basis that will allow me to employ this kind of cascaded processing strategy? 

Again, I appreciate your prompt replies and I am most impressed with TradersStudio!

You can then create a session using this new data with the extended fields. We can cut the the processing by looking at our selected filter fields and doing only the required processing.

      Here is a question, You are short a stock it meets your screen, which is a price above 10.00, what happens if the price drops below 10.00, if you don't check the stock , you will not get out of the trade ?. This is an issue with the classic filter approach.

  These are the type of issue we need to address in developing this application.

• I could create a session that extends the data fields with a simple "0/1" flag based upon my selection filters. 
  
• Having done this once, there would be no need to run the session in subsequent passes unless of course I desired to change my filter critieria
• If I am using CSI data, do I therefore have to export to a CSV format to extend the data model, or does CSI allow such a thing? 
  
• How difficult is this to maintain when using the unadjusted/splitadjusted/dividend adjusted model you present in the tutorial?