The purpose of pipeline is to dynamically filter stocks based on changing conditions - you can't try to add stocks that shouldn't be in there inside the pipeline since the filters are defined once at the start of the program.

Instead, if you wanted to add some trading logic regarding the stock that you currently hold a position in but isn't selected by the pipeline you should loop over your context.portfolio.positions variable which automatically tracks your open positions. You should most likely do this inside your rebalance function.

Hope this helps :)!

@Matthew,

Thank you for the reply, I appreciate it.

I do see how the pipeline is a dynamic tool. However, my concern is that if I open a position (based on pipeline output), and I decide what to do with this position (increase or close it out) based on statistics output from the pipeline (like a factor) what am I supposed to do if that instrument is dropped from the pipeline at the beginning of the next month? I see it as I'm more or less "flying blind" with my position. I have some logic in place to close a position out if I hold it for X number of days... but what about other strategies that might look to hold a position for a much longer time frame?

As an aside... my understanding is that it's not possible to set a pipeline screen that statically selects a set of instruments (please correct me if I am wrong). What if I am developing a strategy that I only want to test against SPY? I understand that the pipeline allows me to discover that this strategy might work well against more than SPY, but if the only solution is to calculate all of my factors for the entire 8000+ instruments on Quantopian, and then filter that output in before_trading_start(), I feel like that's a decent waste of computation, especially if your factors are expensive to calculate. Am I incorrect here, or is this the only way to get a pipeline to calculate factors for a fixed set of securities? (this last paragraph was fairly off-topic, I'm sorry)

Don't mean to whine about the platform, I think it's amazing. Just trying to learn! Thank you for the help.

I think I understand your problem. You want to use the pipeline to filter the stocks, but you also want to see data of stocks which don't pass the filter.

The very purpose of pipeline is stock selection is universe filtering, and conveniently not seeing stocks which don't pass the filter.

So if you wanted to see data from stocks that don't pass the filter, then don't filter the pipeline output

Instead, you can add a column in your pipeline which tracks the value of whether the security passes the filter (true or false), and not actually filter the pipeline output (which is set by screen).

Then, for stocks in your existing portfolio, you could check the value of the various factors that were computed in your pipeline, since they haven't been filtered out.

There is something similar to this in this tutorial, where he sets a column named longs and uses the truth value of that column to determine whether to go long on the stock. However, in your case, you wouldn't want to set the pipeline screen (longs | shorts), since you might want to see stocks that don't fit either criteria.

You can use pipeline to statically select instruments although it's not recommended. You can do something like

sym = morningstar.share_class_reference.symbol  
security_filter = sym.element_of(list of stocks you want)  

and screen on security_filter, which will compute pipeline values only for stocks in your list.

Matthew,

Yes! The way you re-worded the problem makes things so much clearer. Thank you! I can use the pipeline to select a universe while still getting data for that universe and any other instruments which may be existing in my portfolio.

I think I was also getting too caught up with trying to optimize pipeline output.

Regardless, thank you very much!

Glad to help :)