Skip to content

Latest commit

 

History

History
141 lines (114 loc) · 6.15 KB

migration.md

File metadata and controls

141 lines (114 loc) · 6.15 KB
Raw

Migrate your Pipeline from Quantopian

pipeline-live helps you run your algorithm outside of the Quantopian. Although this project is an independent effort to provide the Pipeline API using public/private data, this document is to describe the common practices around how to migrate your pipeline code from the Quantopian environment.

Along with these practices, you can migrate your Algorithm API from Quantopian using pylivetrader, and pylivetrader can run the pipeline object from this package.

USEquityPricing

The most important class to think about first is the USEquityPricing class and it is well covered by pipeline_live.data.alpaca.pricing.USEquityPricing class. This class gets the market-wide daily price data (OHLCV) up to the previous day from Alpaca data API.

Factors

In order to use many of the builtin factors with this price data loader, you need to use pipeline_live.data.alpaca.factors package which has all the builtin factor classes ported from zipline. Use of the Alpaca data API requires an Alpaca account, which you can sign up for here.

Once you have an Alpaca account, you will need to store your account info from their dashboard as environment variables. You can find information about how to do so on this documentation page.

To use the Alpaca factors, import them from pipeline_live.data.alpaca.factors. For example, if you have these lines on Quantopian,

from quantopian.pipeline.factors import (
    AverageDollarVolume, SimpleMovingAverage,
)
from quantopian.pipeline.data.builtin import USEquityPricing

you can rewrite it to something like this.

from pipeline_live.data.alpaca.factors import (
    AverageDollarVolume, SimpleMovingAverage,
)
from pipeline_live.data.alpaca.pricing import USEquityPricing

Of course, the builtin factor classes in the original zipline are mostly pure functions and take inputs explicitly, so if you give the correct ones, they also work with this USEquityPricing.

from zipline.pipeline.factors import AverageDollarVolume
from pipeline_live.data.alpaca.pricing import USEquityPricing

dollar_volume = AverageDollarVolume(
    inputs=[USEquityPricing.close, USEquityPricing.volume],
    window_length=20,
)

The only difference in the factor classes in pipeline_live.data.alpaca.factors is that some of the classes have Alpaca's USEquityPricing as the default inputs, so you don't need to explicitly specify it.

Fundamentals

The Quantopian platform allows you to retrieve various proprietary data sources through pipeline, including Morningstar fundamentals. While the intention of this pipline-live library is to add more such proprietary data sources, the alternative at the moment is IEX. There are two main dataset classes are builtin in this library, IEXCompany and IEXKeyStats. Those both belong to the pipeline_live.data.iex.fundamentals package.

Please note that, in order to use the IEX API data, you will need to sign up for an IEX Cloud account here and set your IEX Cloud token in the IEX_TOKEN environment variable. IEX limits your API messages per month. In order to avoid running over your message quota, please make sure that you filter your stock universe as much as possible before using IEX API data. If you wish to use IEX data to frequently filter a larger set of symbols, you may need to upgrade your IEX Cloud account.

IEXCompany

This dataset class maps the basic stock information from the Company API. If your Quantopian algorithm is using symbol filtering from Morningstar, you can reference the symbol field from this class.

    # not_wi = ~morningstar.share_class_reference.symbol.latest.endswith('.WI')
    not_wi = ~IEXCompany.symbol.latest.endswith('.WI')

Also you can filter out Limited Partners using the companyName field.

    # not_lp_name = ~morningstar.company_reference.standard_name.latest.matches('.* L[. ]?P.?$')
    not_lp_name = ~IEXCompany.companyName.latest.matches('.* L[. ]?P.?$')

The sector and industry fields are good to use for classifiers as well.

IEXKeyStats

This dataset class maps the detailed statistics for the stock from the Key Stats API. The most common use case of this class is the marketcap field.

    # market_cap = morningstar.valuation.market_cap >= 100e6
    market_cap = IEXKeyStats.marketcap.latest >= 100e6

Note, all the fields under IEXKeyStats fields needs .latest access to use the value, unlike the Quantopian's morningstar package.

While IEX's API provides three last quarter financial reports, currently pipeline-live does not provide historical view of this data through the pipeline interface.

Primary Share

Many algorithms developed in the Quantopian platform uses the IsPrimaryShare function to perform base filter. While this value is unique to Morningstar and IEX does not provide this value, something similar can be filtered, at least the following criteria.

  • Has valid revenue value (excluding funds, non-corporate type of shares)
  • Has the biggest marketcap/dollar volume within the same company name (choosing one of the shares between pairs such as GOOG/GOOGL, BRK.A/BRK.B)

The pipeline_live.data.iex package currently does not provide this logic as a function, until we confirm the result is good enough for real uses, but you can built your own function which implements something similar.

Within this library, you have access to Polygon fundamentals, which have different set of stock info/details. If you have API key for Polygon, you may want to look at the pipeline_live.data.polygon.filters.IsPrimaryShareEmulation class for the replacement, too.

ADR

While IEX Company API has a field called issueType and the API document indicates this field can use for ADR check, this value has been not very accurate, as of writing. If you have access to Polygon, you can check out PolygonCompany.country field to filter out non-US companies.