Historically, alpha factors used a single input and simple heuristics, thresholds or quantile cutoffs to identify buy or sell signals. ML has proven quite effective in extracting signals from a more diverse and much larger set of input data, including other alpha factors based on the analysis of historical patterns. As a result, algorithmic trading strategies today leverage a large number of alpha signals, many of which may be weak individually but can yield reliable predictions when combined with other model-driven or traditional factors by an ML algorithm.

The open source zipline library is an event-driven backtesting system maintained and used in production by the crowd-sourced quantitative investment fund Quantopian (https://www.quantopian.com/) to facilitate algorithm-development and live-trading. It automates the algorithm's reaction to trade events and provides it with current and historical point-in-time data that avoids look-ahead bias.

You can use it offline in conjunction with data bundles to research and evaluate alpha factors. When using it on the Quantopian platform, you will get access to a wider set of fundamental and alternative data. We will also demonstrate the Quantopian research environment in this chapter, and the backtesting IDE in the next chapter. The code for this section is in the 01_factor_research_evaluation sub-directory of the GitHub repo folder for this chapter.

After installation and before executing the first algorithm, you need to ingest a data bundle that by default consists of Quandl's community-maintained data on stock prices, dividends and splits for 3,000 US publicly-traded companies. You need a Quandl API key to run the following code that stores the data in your home folder under ~/.zipline/data/<bundle>:

$ QUANDL_API_KEY=<yourkey> zipline ingest [-b <bundle>]