Yes, the specific returns, along with a number of attributed returns, can easily be retrieved as a pandas dataframe from the backtest AlgorithmResult object. There's documentation on this attributed_factor_returns attribute in the docs (
https://www.quantopian.com/docs/api-reference/research-api-reference#quantopian.research.AlgorithmResult.attributed_factor_returns)

The first step is to get a reference to this object using the get_backtest method. This returns a AlgorithmResult object from which one can then reference the various attributes. For example, if one wanted a dataframe of the returns, something like this should work (make sure you use your own backtest ID) .

# replace with your backtest ID  
my_backtest_id = '5dd3f9dc2c781c4bfeaae1b0'  
bt = get_backtest(my_backtest_id)  
attributed_returns = bt.attributed_factor_returns  
specific_returns = attributed_returns.specific_returns

The above will fetch a dataframe with the attributed returns as columns. One of the columns is 'specific_returns'. The dataframe can be copied, manipulated, plotted, and analyzed as desired. There is a complete list of algorithm result attributes and methods in the docs (https://www.quantopian.com/docs/api-reference/research-api-reference#algorithm-results).

Hope that helps.

Hi Dan,

What if I don’t have a backtest ID and am just running a pipeline factor in Research? Can I still get the specific returns from the factor?

@Joakim Alphalens currently doesn't have methods to extract specific returns. There is the method alphalens.performance.factor_alpha_beta (https://quantopian.github.io/alphalens/alphalens.html#alphalens.performance.factor_alpha_beta) which can extract alpha. This alpha is the excess returns above the market (beta). It's not though, returns not attributed to all common exposures.

One could make a simple algo which uses a single factor. This algo then could be backtested and then specific returns analyzed.