Getting Fundamentals Example¶

#Import basic pipeline stuff
from quantopian.pipeline import Pipeline, CustomFactor
from quantopian.research import run_pipeline

from quantopian.pipeline.filters import StaticAssets, Q1500US

import pandas as pd

# First thing is to import the morningstar dataset
# Most fundamentals are here
import quantopian.pipeline.data.morningstar as mstar

# Find the specific data fields you want
# If you want the latest data then simply create factors using the .latest method
filed_date = mstar.financial_statement_filing.file_date.latest
as_of_date = mstar.financial_statement_filing.period_ending_date.latest
form = mstar.financial_statement_filing.form_type.latest

free_cash_flow = mstar.cash_flow_statement.free_cash_flow.latest
sales = mstar.income_statement.total_revenue.latest

# If you need custom ratios or formulas they can typically be created by manipulating
# the latest factors like this
fcf_margin = free_cash_flow / sales

# If you are looking for historical fundamentals you will need to create a simply custom factor
# Long lookbacks can take a long time. Best to use the 'mask' feature to reduce universe size.
class Historical_Data(CustomFactor):
    # Must explicitly enter input and window_length
    # These are the data and lookback length respectively

    def compute(self, today, assets, out, data):
        # returns the earliest data in the dataset
        out[:] = data[0]  

# Create a factor for cash flow 1 year ago. There are typically 252 trading days in a year.
free_cash_flow_yr_ago = Historical_Data(inputs=[mstar.cash_flow_statement.free_cash_flow], window_length=252)
fcf_annual_increase = free_cash_flow/free_cash_flow_yr_ago

# Create a pipline and add any factors as columns
pipe = Pipeline()
pipe.add(filed_date, 'filed_date' )
pipe.add(as_of_date, 'as_of_date' )
pipe.add(form, 'form' )
pipe.add(free_cash_flow, 'free_cash_flow' )
pipe.add(free_cash_flow_yr_ago, 'free_cash_flow_yr_ago' )
pipe.add(fcf_annual_increase, 'fcf_annual_increase' )


# Print out the pipeline results with results for each security (sorted by yield)
# Notice this is all securities for a single day (this is what a piepline returns in the algo environment)
results = run_pipeline(pipe, '2016-3-27', '2016-3-27')
results.sort_values('fcf_annual_increase', ascending=False)

# Note that there are a lot of NaNs. These can come from non-stock securities
# Best to filter those out. An easy way to get the largest 1500 stocks is to use
# the built in Q1500US filter
q1500_filter = Q1500US() 

# Lets add that filter to our pipeline and re-run to see the results
# Notice that there are still some weird data

pipe.set_screen(q1500_filter)
results = run_pipeline(pipe, '2016-3-27', '2016-3-27')
results.sort_values('fcf_annual_increase', ascending=False)

# Lets see what the data looks like over time now.
# Remember that the results above were for a single day.
# Best to do this with a single stock so lets create a filter for AAPL

aapl_filter = StaticAssets(symbols(['AAPL']))
pipe.set_screen(aapl_filter, overwrite=True)
results = run_pipeline(pipe, '2015-10-27', '2016-3-27')
results