Is it possible to get 5-year average Dividend Yield using zipline in QuantRocket?

Technical Help
#1

I am considering to purchase quantrocket software. But I have a problem as follows.
I have a rebalance strategy , which rebalances a portfolio every week.
On the rebalance day, for each stock, I need it's average Dividend Yield within last 5 years.That means we need its every day's Dividend Yield within last 5 years, so that we can compute this 5-year average.
I wonder if it's possible to get this 5-year average Dividend Yield using zipline in QuantRocket?

In Sharadar's fundationmental dataset, I see DIVYIELD (Dividend Yield measures the ratio between a company's [DPS] and its [PRICE]),
I guess we can get 5-year average Dividend Yield through this factor in zipline. And I also want to get the quantiles of Dividend Yield within last 5 years, is it possible using Sharadar's fundationmental dataset to get it?
Could anyone assure this?

Thanks in advance.

#2

This can be done in Pipeline using the period_offset parameter:

from zipline.pipeline.data.sharadar import Fundamentals

divyield_0 = Fundamentals.slice(dimension="ARY", period_offset=0).DIVYIELD.latest
divyield_1 = Fundamentals.slice(dimension="ARY", period_offset=-1).DIVYIELD.latest
divyield_2 = Fundamentals.slice(dimension="ARY", period_offset=-2).DIVYIELD.latest
divyield_3 = Fundamentals.slice(dimension="ARY", period_offset=-3).DIVYIELD.latest
divyield_4 = Fundamentals.slice(dimension="ARY", period_offset=-4).DIVYIELD.latest

divyield_avg = (divyield_0 + divyield_1 + divyield_2 + divyield_3 + divyield_4) / 5

divyield_avg_quantiles = divyield_avg.quantiles(5)

In QuantRocket version 2.9 (not yet released) this will become even easier with some new built-in factors for working with periodic data like Sharadar fundamentals:

from zipline.pipeline.periodic import PeriodicAverage

divyield = Fundamentals.slice('ARY').DIVYIELD
divyield_avg = PeriodicAverage(divyield, window_length=5)
...
#3

Thank you very much @Brian. Maybe I didn't express my scenario clearly.
By saying 5-year average Dividend Yield, it doesn't mean the average of just 5 data points. to put it another way, for example, I need average Dividend Yield for last 450 days. Then it means we need 450 data points to compute this average. I wonder how to get it in zipline backtesting enviroment (not research enviroment)?

By definition, DIVYIELD (Dividend Yield measures the ratio between a company's [DPS] and its [PRICE]). It varies every day, because the price varies every day.

#4

1 After some study, I guess the DIVYIELD doesn't change every day, it is based on the price of date DATAKEY. am I right? and how to get this DATAKEY in zipline?

2 So, in order to get everyday dividendRatio in zipline pipline, I do the folloing:

DIVYIELD = sharadar.Fundamentals.slice(dimension='ART', period_offset=0).DIVYIELD.latest
DPS = sharadar.Fundamentals.slice(dimension='ART', period_offset=0).DPS.latest

close = USEquityPricing.close.latest

dividendRatio = DPS / close
pipe = Pipeline(
    columns={
        'DIVYIELD': DIVYIELD,
        'DPS': DPS,
        'close': close,
        'dividendRatio':dividendRatio,     
    },
 
)
res = run_pipeline(pipe, start_date='2016-01-04', end_date='2023-02-23', bundle='usstock-1d-bundle')
print("There are %d assets in this universe." % len(res))

res.head # print 10 constituents

I can get dividendRatio for each day now. Is this solution correct?
next step, I will use this dividendRatios to compute e.g, 400 day average.

Can quantrocket introduce the dynamic daily pe in sharadar?
#5

This is the right idea but it will yield incorrect results when stock splits have occurred because the fundamental fields like DPS are fully split-adjusted while the bundle data is point-in-time split-adjusted, that is, only adjusted for splits that have occurred up to the current end date of the running Pipeline. This distinction is explained in the docs.

A better solution that avoids this problem is to collect the history db (usstock or sharadar) and import the close price into Pipeline as a custom database. The history databases are fully split-adjusted, like the fundamental data, so you can combine them without a problem:

from zipline.pipeline.data.db import Database, Column

class FullyAdjustedEquityPricing(Database):

    CODE = "sharadar-stk-1d"
    Close = Column(float)

dividendRatio = DPS / FullyAdjustedEquityPricing.Close.latest
#6

good.
So, is the following the correct method to compute dynamic PB(Price to Book Value) ?

    class FullyAdjustedEquityPricing(Database):
        CODE = "sharadar-us-stk-1d" 
        Close = Column(float) 

    equityq = sharadar.Fundamentals.slice(dimension='ARQ', period_offset=0).EQUITY.latest
    SHARESBAS = sharadar.Fundamentals.slice(dimension='ARQ', period_offset=0).SHARESBAS.latest

    equityPerShare = equityq/SHARESBAS    
    pb = FullyAdjustedEquityPricing.Close/equityPerShare

    return Pipeline(
        columns={     
            'pb':pb, 
        },  
    )