Good question! (and welcome)

What is the difference between Returns and pct_change?

First off 'Returns' is a factor (https://www.quantopian.com/help#built-in-factors) while 'pct_change' is a pandas dataframe method (https://pandas.pydata.org/pandas-docs/version/0.18/generated/pandas.DataFrame.pct_change.html). Let's see what that means.

Factors are used to define the columns which are returned from a pipeline. Note the term 'define'. Returns itself doesn't directly manipulate any data - just set's up the mechanism to fetch data and perform any calcs. The code below will create a pipeline with a single column for returns.

pipe = Pipeline(  
    columns = {'returns': Returns(inputs=[USEquityPricing.close], window_length=2) }  
    )

Pandas methods, on the other hand, directly manipulate data in a dataframe. The output of a pipeline is a pandas dataframe so any of the pandas methods can manipulate the output of a pipeline but only after it has been run. The code below will create a new column 'percent_change' in the dataframe returned by a pipeline.

result['percent_change'] = result.groupby(level='stock').close_price.pct_change()

In the end one will get the exact same number in a column of a dataframe.

There are several advantages to getting returns using the Returns factor however. First, the pct_change approach will leave a NaN value in the first row while Returns will not. Second there is a bit cleaner way to get log returns by using the Returns factor. Finally, one can create filters and other factors going the factor approach which can't be done if calculating the pct_change once pipeline is run. See the attached notebook for a bit of explanation of each of these.

Hope that clears it up a bit.

(Not to confuse things more but there is also a data fetch method called returns (notice the small r) https://www.quantopian.com/help#quantopian_research_returns. Assuming that's not what the question was asking? )

Thanks, Dan! That's exactly what I want to know.

Very helpful, thank you!