Quantitative investment strategy: several common Python back-testing frameworks (libraries)

Quantitative investment strategy: several common Python back-testing frameworks (libraries)

Original address: http://blog.csdn.net/lawme/article/details/51454237

This article is reproduced in the article. This time in the study of quantitative strategy direction, studied zipline for some time, but the follow-up found he only support U.S. stocks, collect quantitative strategy articles, reproduced in the blog.

The quantitative trading strategy must be backtesting before real trading. Here, we evaluate the commonly used Python backtesting framework (library). The scale of the evaluation includes the scope of use (backtesting, virtual trading, real trading), ease of use (good structure, document integrity) and extensibility (fast, simple to use, compatible with other framework libraries).

• Zipline: Event-driven backtesting framework. Quantopian is using it.
  • Zipline has a large community, complete documentation, strong support for the well-known brokerage firm interactive Broker (IB), integrated pandas, clear grammar, easy to learn and master.
  • There are a lot of routines examples. If you are primarily trading US securities, it is the best option. Quantopian allows backtesting, sharing, and discussing trading strategies in their communities.
  • However, according to our experience, the speed ofZipline is very slow. This is its largest short plate. Quantopian some countermeasures, such as running in parallel in the cloud. If you are interested, you can read this blog post.
  • Zipline seems difficult to use local data files and non-American data.
  • It is difficult to use for different kinds of financial assets.

• Pyalgotrade: is also an event-driven backtesting framework that supports both virtual and real-time transactions. The document is complete and integrates the Ta-lib (Technical Analysis Library). In terms of speed and flexibility, it is stronger than Zipline . However, one of its major mishap is the modules and objects that do not support Pandas.

• pybacktest: It is a very simple and lightweight method of backtesting with vector data. May 21, 2015, the project has a resurgence of signs.

• Tradingwithpython: This jev Kuznetsov expands pybacktest to form its own backtesting program. This library seems to have been updated in February 2015. However, the relevant documentation and course prices are $395.

• Other Items: ultra-finance

	Zipline	Pyalgotrade	Tradingwithpython	Pybacktest
Type	Event-driven	Event-driven	Vector processing	Vector processing
Community	Larger	So so	No	No
Cloud computing	Quantopian	No	No	No
Support IB	Is	Whether	Whether	Whether
Data source	Yahoo, Google, NinjaTrader	Yahoo, Google, NinjaTrader, Xignite, Bitstamp provide data in real time
Document	Complete	Complete	$395	Rarely
Events can be customized	Is	Is
Speed	Slow	Fast
Support Pandas	Is	Whether	Is	Is
Trading Calendar	Is	Whether	Whether	Whether
Support Ta-lib	Is	Is	Is
Suitable for	For US securities Trading only	Real Trading Virtual Disk Trading	Virtual Disk test Transaction	Virtual Disk test Transaction

Comparison of Zipline and Pyalgotrade score

 	zipline	pyalgotrade	description
Virtual disk trading	"	?  ? ?	zipline does not seem to work with non-US data and local data, and Pyalgotrade can use any type of data
Real trading	??	?	Both are good, but Quantpian's cloud computing programming is good
flexibility	??	??	pyalgotrade supports a variety of advanced orders and has more business events. Zipline provides a simple slippage mode
speed	?	??	zipline slower than Pyalgotrade
Ease of use	???	?	pyalgotrade not supported pandas

Quantitative investment strategy: several common Python backtesting frameworks (libraries)