Pipeline data is split and dividend adjusted. It's just adjusted as of each pipeline date. The goal of pipeline is to provide point-in-time data which is what a trader (or algorithm) would have seen on a specific pipeline date. Providing forward adjusted prices would introduce forward lookahead bias. If one wants adjusted prices as of a specific day then one can use the get_pricing or similar method. Check out the documentation https://www.quantopian.com/docs/api-reference/research-api-reference#price-and-volume-data.

There are a couple of posts which go into more detail on this (and interestingly use AAPL as an example too).
https://www.quantopian.com/posts/research-updates-get-pricing-and-jupyter-notebook-upgrade
https://www.quantopian.com/posts/usequitypricing-and-get-pricing-data-difference
https://www.quantopian.com/posts/wrong-stock-price

Thanks Dan.

I would have thought that pipeline would adjusted for split based on the end_date parameter.

For instance:

pipe = Pipeline(  
    columns={  
        'close': USEquityPricing.close.latest,  
    },  
    domain=US_EQUITIES,  
)

# filter  
symbol = StaticAssets(symbols(['AAPL']))  
pipe.set_screen(symbol)  
df = run_pipeline(pipe, '2005-02-20', '2005-03-05')  

The result shows no adjustment:
2005-02-22 00:00:00+00:00 Equity(24 [AAPL]) 86.740
2005-02-23 00:00:00+00:00 Equity(24 [AAPL]) 85.380
2005-02-24 00:00:00+00:00 Equity(24 [AAPL]) 88.220
2005-02-25 00:00:00+00:00 Equity(24 [AAPL]) 88.900
2005-02-28 00:00:00+00:00 Equity(24 [AAPL]) 44.485
2005-03-01 00:00:00+00:00 Equity(24 [AAPL]) 44.780
2005-03-02 00:00:00+00:00 Equity(24 [AAPL]) 44.430
2005-03-03 00:00:00+00:00 Equity(24 [AAPL]) 44.130
2005-03-04 00:00:00+00:00 Equity(24 [AAPL]) 41.800
2005-03-07 00:00:00+00:00 Equity(24 [AAPL]) 42.800

Does that mean my CustomFactor is using a non-adjusted price all this time?

I can switch to get_pricing in my research. But I need the universe selection capability of the pipeline. Is this possible?

Ok correction, after some experiments I got it. The price is adjusted during CustomFactor: on 2005-2-25, when we look at value[-1] it is at 88.9. But on the next tick on 2005-2-28, if we look back to 2005-2-25 at value[-2], it is adjusted accordingly to 44.45.

Following is the custom factor code that I used for the experiment in case someone is interested.

class get_adj(CustomFactor):  
    window_length = 2  
    def compute(self, today, assets, out, value):  
        out[:] = value[-2]  

So, if we use the price in CustomFactor, it is safe. My problem arises when I try to create new columns from pipeline result (pandas dataframe). If I stick to just CustomFactor this should be fine.

Although, having the universe functionality in get_pricing does come in handy.