The pipeline concept can be a bit difficult to wrap one's head around at first. But, once over the initial 'hump', it's pretty straightforward. Have no fear.

Also, it doesn't help that the the word 'pipeline' is used in other areas where it doesn't completely match with the Quantopian definition. The scikit-learn link listed above is an example. It's conceptually similar (ie a pipeline is 'thing' which performs sequential tasks) but isn't really the same other than the name.

Let's start with the imports. When one imports the following

from quantopian.pipeline.data import EquityPricing, factset

the two classes EquityPricing and factset are DataSets. They are definitions for how to connect to the correct field in the data database. They aren't the data itself just the connection info. So, doing something like this

df = USEquities  

doesn't assign any data to 'df' rather just info on how to get the data. To get a list (or more precisely a dataframe) of the data, one needs to execute the pipeline using the run_pipeline method. The result of that is the data. Maybe read through the docs (https://www.quantopian.com/help#base-classes) for the definition of BoundColumn.

All that said, there is another post which may make it more clear. Take a look at this (https://www.quantopian.com/posts/need-help-with-pipeline-just-a-beginner ).

Hope that helps. Good luck.

It really helped, thanks! Another question then, if you have "Quantopian own definition of pipelines", what else is just "Quantopian defined"? Reading, I know that the factors used in pipelines are also "Quantopian defined". What else? Because I don't want to switch to python in my job and have the surprise something I learned here It's not going to work anywhere else. I would appreciate knowing before-hand.

The pipeline uses auxiliary objects like factors, filters and classifiers, as well as datasets, which are incorporated in the quantopian API. The API also makes use of Equity objects, which are used as most commonly as indices in the pandas MultiIndex for the pandas DataFrame and Series objects, as well as futures and continuous futures objects for futures based strategies. Overall, they are nothing more than summary classes which gather the necessary information for the respective asset class. Refer to the API documentation, it is explained very well there. On another note, the output of the quantopian pipeline is a pandas DataFrame, which makes it easy to manipulate both in research and in the backtest environment. The idea behind the pipeline is to generalize the operations needed in an algorithmic strategy, so if you have the data yourself you should be able to achieve approximately the same results manually, however, it would presumably take more effort than using the pipeline provided here.